Climate Modelling, Climate Prediction and Model Validation
W.L. GATES, J.F.B. MITCHELL, G.J. BOER, U. CUBASCH, V.P. MELESHKO
Contributors: D. Anderson; W. Broecker; D. Cariolle; H. Cattle; R.D. Cess; F. Giorgi; M.I. Hojfert; B.G. Hunt; A. Kitoh; P. Lemke; H. Le Treut; R.S. Lindzen; S. Manabe; B.J. McAvaney; L. Mearns; G.A. Meehl; J.M. Murphy; T.N. Palmer; A.B. Pittock; K. Puri; D.A. Randall; D. Rind; PR. Rowntree; M.E. Schlesinger; C.A. Senior; I.H. Simmonds; R. Stoujfer; S. Tibaldi; T. Tokioka; G. Visconti; J.E. Walsh; W.-C. Wang; D. Webb
CONTENTS
Executive Summary 101 B4.2 Diurnal Cycle 119 B4.3 Day-to-Day Variability 119 Bl Introduction 103 B4.4 Extreme Events 119 B2 Advances in Modelling Climate Change due to B4.5 Blocking and Storm Activity 119 Increased Greenhouse Gases 103 B4.6 Inlra-Seasonal Variability 120 B2.1 Introduction 103 B4.7 Interannual Variability 120 B2.2 New Transient Results from Coupled B4.8 ENSO and Monsoons 121 Atmosphere-Ocean GCMs 103 B4.9 Decadal Variability 122 B2.3 New Equilibrium Results from Atmospheric B5 Advances in Modelling the Oceans and Sea-Ice 122 GCMs and Mixed-Layer Ocean Models 108 B5.1 Introduction 122 B2.4 Regional Climate Simulations 112 B5.2 Eddy-Resolving Models 123 B3 Advances in Analysis of Climate Feedbacks and B5.3 Thermohaline Circulation 124 Sensitivity 114 B5.4 Sea-Ice Models 125 B3.1 Introduction 114 B3.2 Water Vapour Feedback 114 B6 Advances in Model Validation 126 B6.1 Introduction 126 B3.3 Cloud Feedback 116 B6.2 Systematic Errors and Model Intercomparison 126 B3.4 Surface Albedo Feedback 117 B6.3 Data for Model Validation 129 B3.5 Climate Sensitivity 118 B4 Advances in Modelling Atmospheric Variability 118 References 129 B4.1 Introduction 119
EXECUTIVE SUMMARY
Coupled Model Experiments the equilibrium sensitivity to doubled CO2 from the range The new transient climate simulations with coupled of 1.5 to 4.5°C as given by IPCC 1990. atmosphere-ocean general circulation models (GCMs) • There is no compelling evidence that water vapour generally confirm the findings of Section 6 in the 1990 feedback is anything other than the positive feedback it has Intergovernmental Panel on Climate Change report (IPCC, been conventionally thought to be, although there may be 1990), although the number of coupled model simulations difficulties with the treatments of upper-level water vapour is still small. in current models. • There is broad agreement among the four current models • The effects of clouds remain a major area of uncertainty in in the simulated large-scale patterns of change and in their the modelling of climate change. While the treatment of temporal evolution. clouds in GCMs is becoming more complex, a clear • The large-scale patierns of temperature change remain understanding of the consequences of different cloud essentially fixed with time, and they become more evident parametrizations has not yet emerged. with longer averaging intervals and as the simulations progress. Atmospheric Variability • The large-scale patterns of change are similar to those • Model experiments with doubled CO2 give no clear obtained in comparable equilibrium experiments except indication of a systematic change in the variability of that the warming is retarded in the southern high latitude temperature on daily to interannual time-scales, while ocean and the northern North Atlantic Ocean where deep changes of variability for other climate features appear to water is formed. be regionally (and possibly model) dependent. All but one of the models show slow initial warming (which may be an artefact of the experimental design) Ocean Modelling followed by a nearly linear trend of approximately 0.3°C • Results from eddy-resolving ocean models show a broadly per decade. realistic portrayal of oceanic variability, although the All models simulate a peak-to-trough natural variability of climatic role of eddies remains unclear. about 0.3°C in global surface air temperature on decadal Ocean-only GCMs show considerable sensitivity of the time-scales. thermohaline circulation on decadal and longer time-scales • The rate of sea level rise due to thermal expansion to changes in the surface fresh-water flux, although increases with time to between 2 and 4cm/decade at the coupled models may be less sensitive. time of doubling of equivalent CO2. • Sea-ice dynamics may play an important role in the freezing process, and should therefore be included in Regional Changes models for the simulation of climate change under • Although confidence in the regional changes simulated by increased CO2 GCMs remains low, progress in the simulation of regional climate is being obtained with both statistical and one-way Model Validation nested model techniques. In both cases the quality of the • There have been improvements in the accuracy of large-scale flow provided by the GCM is critical. individual atmospheric and oceanic GCMs, although the ranges of intermodel error and sensitivity remain large. Climate Feedbacks and Sensitivity • The lack of adequate observational data remains a serious There is no compelling new evidence to warrant changing impediment to climate model improvement.
B Climate Modelling, Climate Prediction & Model Validation 103
Bl Introduction B2 Advances in Modelling Climate Change due to Increased Greenhouse Gases Since the publication of the first IPCC Scientific B2.1 Introduction Assessment of Climate Change (IPCC, 1990) there have Simulation of the climatic response to increases in been significant advances in many areas of climate atmospheric greenhouse gases has continued to dominate research as part of a continuing worldwide acceleration of climate modelling. Preliminary results from new interest in the assessment of possible anthropogenic integrations of coupled global atmosphere-ocean models climate changes. In this section we concentrate on with progressive increases of CO2 show that the patterns of advances in the modelling of climate change due to the transient response are similar to those in an equilibrium increased greenhouse gases, improvements in the analysis response, except over the high-latitude southern ocean and of climate processes and feedbacks, and on advances in northern North Atlantic ocean; here the delayed warming climate model validation that were not available to the has highlighted the critical role of the oceanic IPCC in early 1990. This section is thus intended as an thermohaline circulation. Computing limitations have update to selected portions of the 1990 assessment rather continued to restrict the resolution that can be used in than as a comprehensive revision, and an effort has been GCM simulations of the climate changes due to increased made to keep the discussion both concise and focussed in greenhouse gases, although progress is being made in the accordance with the stringent space limitations placed simulation of regional climate by both statistical upon this supplementary report. techniques and by locally nesting a higher-resolution In Section B2 recent simulations of climate change are model within a global GCM. assessed, with emphasis on results from coupled ocean- atmosphere models. In Section B3, recent research on B2.2 New Transient Results from Coupled Atmosphere- modelling climate feedbacks is discussed, and the IPCC Ocean GCMs 1990 estimates of chmate sensitivity are reviewed. The At the time of the 1990 IPCC report, preliminary results simulation of atmospheric variability and its changes due were available from only two coupled model integrations to increased atmospheric CO2 are discussed in Section B4 with transient CO2, namely those made at NCAR while developments in ocean and sea-ice modelling are (Washington and Meehl, 1989) and at GFDL (Stouffer et presented in Section B5. Finally, Section B6 discusses al, 1989). Of these only the GFDL integration had been advances in climate model validation. carried to the point of CO2 doubling (which occurred after
Table Bl: Summary of transient CO2 experiments with coupled ocean-atmosphere GCMs
GFDL MPI NCAR UKMO
AGCM R15L9 T21 L19 R15L9 2.5° X 3.75° LI 1 OGCM 4.5° X 3.75° L12 4°L11 5°L4 2.5° X 3.75° LI7 Features no diurnal cycle. prognostic CLW, quasi- prognostic CLW, isopycnal ocean geostrophic ocean isopycnal ocean diffusion diffusion Flux adjustment seasonal, heat, fresh seasonal, heat, fresh none seasonal, heat, fresh water water, wind stress water
Control CO2 (ppm) 300 390 t 330 323
CO2 (t) l%yr-' IPCCa&d(1990 1% yr' l%yr-l (compound) Scenarios), 2XCO2 (linear) (compound) Length (yr) 100 100 60 75 CO2 doubling time (yr) 70 60 (IPCCa) 100 70 Warming (°C) at COj 2.3 1.3 2.3 1.7 doubling (projected)
2xC02 sensitivity (°C) 4.0 2.6 4.5 2.7 tt (with mixed layer ocean)
L - number of vertical levels; R - number of spectral waves (rhomboidal truncation); T - no spectral waves (triangular truncation), t - equivalent CO2 value for trace gases ++ - estimate from low resolution experiment 104 Climate Modelling, Climate Prediction & Model Validation B approximately 70 years as a result of a 1% per year compound increase of CO2). Here we present further 4 resuhs from the transient CO2 experiments with both the • GFDL GFDL and NCAR coupled models, along with preliminary • MPI results from new transient CO2 integrations recently — — UKMO completed at the Max-Planck-Institute for Meteorology NCAR 0) —— IPCC A (MPI) in Hamburg and at the Hadley Centre of the United S 1 - Kingdom Meteorological Office (UKMO). A fuller ^ . E description and intercomparison of these results is being ^ 0- prepared (WCRP, 1992). -1 A summary of the new transient results is presented in 0 20 40 60 80 100 Table Bl. Note that there are differences in the CO2 Year equivalent doubling times (by a factor of almost two) and Figure Bl: Decadal mean changes in globally averaged surface that the equilibrium sensitivity of the models is different. temperature (°C) in various coupled ocean-atmosphere The UKMO model has the finest horizontal resolution experiments, (see Table Bl). Note that the scenarios employed while the MPI model has the finest vertical resolution in differ from model to model, and that the effect of temperature the atmosphere and includes explicitly the radiative effects drift in the control simulation has been removed. Open boxes = of other trace gases. Three of the four coupled models (see GFDL; solid circles = MPI; triangles with dashed line = UKMO; Table Bl) use a correction or adjustment of the air-sea triangles with dotted line = NCAR (sea temperatures only); solid fluxes so that the C02-induced changes should be line = IPCC 1990 Scenario A "best estimate". interpreted as perturbations around a climate similar to that presently observed. Adjustment of the fluxes also prevents distortion of the CO2 induced perturbation by rapid drift of GCMs is shown in Figure Bl. In spite of the differences in the model climate (since the same corrections are applied the models' parametrizations and in their experimental to the control and anomaly simulations). configurations, all of the models exhibit a number of 1. The new transient climate simulations with coupled similar overall features in their response. Firstly, three of atmosphere-ocean GCMs generally confirm the findings of the models show relatively little warming during the first Section 6 in the 1990 IPCC report, although the number of few decades of the integration rather than a constant rate of coupled model simulations is still small. warming throughout, despite the near constant rate of The globally averaged annual mean increase of surface increase in radiative heating. This so-called "cold start " is air temperature at the time of effective CO2 doubling (70 barely noticeable in the GFDL simulation, but in the years for the GFDL and UJCMO models, 60 years for the UKMO and MPI models the warming is negligible during MPI model with IPCC Scenario A, and 100 years for the the first 2 to 3 decades. This phenomenon is thought to be NCAR model) is between 1.3°C and 2.3°C. These values an artefact of the experimental design and can be are approximately 60% of the models' equilibrium reproduced qualitatively using simplified models warming (where known) with doubled CO2 when run with (Hasselmann et al., 1992; J.M. Murphy, personal simple mixed-layer oceans. These results confirm those communication; see also Hansen et al., 1985) which from the simple models used in IPCC 1990. The lower indicate that the length of delay grows with increasing values of warming compared to the equilibrium model sensitivity (the equilibrium warming for doubling experiments are partly due to the fact that the transient CO2) and with effective heat capacity in the ocean. Until experiments take into account the thermal inertia of the this "cold start" phenomenon is investigated and deep ocean and are therefore not at equilibrium at the time understood, it is not meaningful to match "model time" of effective CO2 doubling. with calendar dates. Secondly, as the increase in CO2 progresses, each model approximates a constant rate of 2. All but one of the models show a limited initial warming warming, in overall agreement with the O.S'C/decade in (which may be an artefact of the experimental design) the IPCC (1990) projections made for the "business-as- followed by a nearly linear trend of approximately 0.3°C usual" CO2 forcing scenario (A) with a simplified per decade. All models simulate a peak-to-trough natural upwelling-diffusion ocean model. Thirdly, all models variability of about 0.3 to 0.4°C in global surface air exhibit variability on interannual, decadal and even longer temperature on decadal time-scales. time-scales, and some models reproduce ENSO-like The evolution of the change of globally averaged effects (Meehl,1990; Lau et al., 1991; J.F.B. Mitchell, surface air temperature (sea surface temperature for the personal communicarion). The peak-to-trough range of NCAR experiment) during the course of the various global surface air temperature intra-decadal natural transient experiments with coupled ocean-atmosphere variability is 0.3 to 0.4°C (see, for example. Figure B2). B Climate Modelling, Climate Prediction & Model Validation 105
3. The rate of sea level rise due to thermal expansion increases with time to between 2 and 4cm/decade at the time of doubling of equivalent COj. The "cold start" leads to a substantial delay in the associated change in global sea level rise due to thermal expansion (Figure B3), and is most pronounced in the UKMO and MPI models. The rate of sea level rise increases during the simulations to 2, 2.5 and 4cm/decade by the time of CO2 doubling in the MPI A (Scenario A), UKMO and GFDL experiments, respectively. The simulated rate of sea level rise at 2030 (the time of doubling of CO2) due to thermal expansion (Scenario "A" best estimate) in the previous IPCC assessment was about 2.5cm/decade. Note that we have not reassessed the contribution to sea level rise from changes in the snow and ice budgets over land.
4. The large-scale patterns of change are similar to those obtained in comparable equilibrium experiments except Figure B2: Temporal variations of the area-averaged deviation that the warming is retarded in the southern high latitude of annual mean surface air temperature (°C) from the ocean and the northern North Atlantic Ocean where deep corresponding 100-year average of the control as simulated by water is formed. the GFDL coupled ocean-atmosphere model for; (a) the globe, The geographical distribution of the change in surface (b) the Northern Hemisphere, and (c) the Southern Hemisphere. air temperature at the approximate time of CO2 doubling in (From Manabe etal., 1991.) the four transient experiments is shown in Figure B4. (For the NCAR model the CO2 increase is only 60%) The same overall characteristics are seen in the distribution of the simulated change of surface air temperature. These characteristics are: (1) the largest warming occurs in the high latitudes of the Northern Hemisphere, (2) relatively uniform warming occurs over the tropical oceans, and (3) a minimum of warming or in some cases cooling occurs over the northern North Atlantic and over the Southern Ocean around Antarctica. Features (1) and (2) are familiar from the earlier studies of the equilibrium warming in response to doubled CO2 with a mixed-layer ocean (see Section B2.3); the high latitude ocean warming minima, on the other hand, are due to the presence of upwelling and deep vertical mixing which increase the effective heat capacity 20 40 60 80 100 and hence the thermal inertia of the ocean locally. Year from start of simulation Figure B3: Decadal mean changes in globally averaged sea level 5. There is broad agreement among the four models in the change (cm) from various coupled ocean-atmosphere GCM simulated large-scale patterns of change and in their experiments (see Table Bl). Open squares = GFDL; solid circles temporal evolution. The large-scale patterns of = MPI (IPCC Scenario "A"); solid diamonds = MPI (IPCC temperature change remain essentially fixed with time, and Scenario "D"); open triangles = UKMO. Note the differences in they become more evident with longer averaging intervals forcing in Table B1, and that the effect of long-term drift in the and as the simulations progress. models has been removed by differencing. A larger warming over land areas compared to that over the oceans is common to all models in both winter and summer. This ocean-land asymmetry contributes to a more rapid warming in the Northern Hemisphere compared to Analysis of the nature of this variability in the coupled that in the Southern Hemisphere. There is a maximum integrations and of the implications for detecting global warming over the Arctic Ocean in winter (Figure B5a) and warming is currently in progress. a minimum in summer (Figure B5b), similar to the equilibrium results. However, the warming and its seasonal .5 IT)
U O E E o o 11 13 c o
^ S 1 - ^ 8 I 8 «
>, am s § I I s CI CTJ o S O ^ O o 2 Si
c <
U
.5 Si > 2 o 2 ^ 00 a 2 S E 00
S3 1
o -a OS o So M> < e U ona
o c o o o 3
60
C _
m C D
CO a. OJ s .s B Climate Modelling, Climate Prediction & Model Validation 107
Figure B5: Distribution of the mean surface air temperature (°C) for (a) DJF, and (b) JJA during years 60-80 of a transient COj simulation with the GFDL model, relative to the 100 year average of the control. (From Manabe et al., 1992.)
variation over the circumpolar ocean in the Southern greenhouse gas concentrations. The cliinate change pattern Hemisphere is considerably less than in the equilibrium becomes more evident as the integration progresses, as models because of deep vertical mixing in the ocean. indicated by the growth in the fraction of variance Precipitation increases in the Northern Hemisphere in high explained by the first EOF of the anomaly experiment latitudes throughout the year, in much of mid-latitudes in (Figure B8). This feature is noticeable in other models (for winter, and in the southwest Asian monsoon (Figure B6a, example, Meehl et al., 1991a) and in other fields such as b). In the Southern Hemisphere, there are precipitation soil moisture (Figure B9) and (to a lesser extent) in the increases along the mid-latitude storm tracks throughout precipitation (Santer et al, 1991; J.F.B. Mitchell, personal the year. In the Northern Hemisphere, winter soil moisture communication). In IPCC 1990 it was assumed that the increases over the mid-latitude continents, while in the regional climate changes would scale linearly with the summer there are many areas of drying; this is also similar global mean temperature changes; while Figure B9 offers to the results obtained by equilibrium models (Figure B7a, some support for this assumption, there is considerable b). interdecadal variability at regional scales in the coupled In at least one simulation (Cubasch et al., 1991) the model simulations, so this scaling remains questionable. progressive change in surface temperature is unrelated to There are a number of caveats to be kept in mind in the pattern of internal variability in the model; the first assessing the results of the transient CO2 simulations with empirical orthogonal function (EOF) of the annual mean coupled ocean-atmosphere models shown here. Chief surface temperature in the control run is uncorrelated with among these is the use (in the GFDL, MPI and UKMO the first EOF from the simulation with increasing models) of adjustments to the oceanic surface fluxes so 108 Climate Modelling, Climate Prediction & Model Validation B
Figure B6: Decadally averaged changes in precipitation (mm/day) for (a) DJF, and (b) JJA around the time of doubling of CO2 in an experiment in which CO2 was increased by 1 %/year in the UKMO model (J.F.B. Mitchell, personal communication). Contours are every 1 mm/day and areas of decrease ai-e stippled.
that the ocean temperature and saUnity remain close to substantially affected by flux adjustment. In addition, present climatology. If flux corrections are not used (as in confidence in the overall validity of the results is raised by the NCAR model), imperfections in the component models the consistency of parallel experiments with the GFDL (and in their interaction) may introduce significant model in which the CO2 undergoes a progressive transient systematic errors in the coupled simulation. Such flux reduction (Manabe et al., 1991,1992). corrections or adjustments represent substantial changes of the fluxes that are exchanged between the component B2.3 New Equilibrium Results from Atmospheric GCMs models and, as shown in Figure BIO, are comparable in and Mixed-Layer Ocean Models magnitude to the atmospheric fluxes. In the case shown, Although increasing attention is being paid to transient they tend to be of opposite sign, i.e., the flux correction simulations with coupled atmosphere-ocean GCMs (see substantially reduces the net effective flux to the ocean. Section B2.2), doubled CO2 experiments with atmospheric The use of such adjustments may distort the models' GCMs and mixed-layer ocean models continue to be of response to small perturbations like those associated with interest since they can be carried to statistical equilibrium increasing CO2. On the other hand, the similarities in the relatively easily and they provide a benchmark for model responses of the NCAR model without flux corrections sensitivity. The results of several new such experiments and in the three models that use flux corrections indicate completed since the publication of the 1990 IPCC that the simulated changes in the models may not be Scientific Assessment (IPCC, 1990) are summarized in B Climate Modelling, Climate Prediction & Model Validation 109
90N
180 150W 120W 90W 60W 30W 0 30E 60E 90E 120E 150E 180
180 150W 120W 90W 60W SOW 0 30E 60E 90E 120E 150E 180
Figure B7: Decadally averaged changes in soil moisture in the MPI model for (a) DJF, and (b) JJA around the time of doubling of effective COj. Contour intervals are every cm. Areas of decrease are shaded. (Cubasch etal, 1991.)
Table B2. In general the simulated globally-averaged cold bias relative to observations of the current climate increases in surface air temperature and precipitation are (Boer et al., 1991a; see Section B5.2), inclusion of the similar to those found earlier (see IPCC, 1990, Table 2.3 trace gases should tend to reduce this systematic error. (a); the differences relative to earlier models can be related Note, however, that the anticipated cooling effect of in some cases to the differences in cloud radiative aerosols (Charlson et al, 1991) is not included in current feedback in the tropics (see Section B3.3)). The largest climate models. changes are found in the one new model (LMD) that The regional climate responses due to the inclusion of includes prognostic cloud liquid water and variable cloud additional greenhouse gases have been examined by Wang optical properties. et al. (1992). In one experiment the trace gases (CO2, New model results suggest that the difference in the CFCs, methane and nitrous oxide) were represented vertical distribution of the radiative forcing has explicitly, while in the second trace gases were represented implications for the simulations of the control climate as by adding the amount of CO2 which gives an equivalent well as for the greenhouse gas-induced warmer climates. increase in radiative heating at the tropopause, i.e., using As demonstrated by Wang et al. (1991b), inclusion of CO2 as a surrogate for the trace gases. When the trace individual trace gases (i.e., not as equivalent CO2) in a gases and CO2 are given values in 2050 according to the control simulation produces a warmer atmosphere, IPCC "business-as-usual" scenario (at which time the especially in the tropical upper troposphere. Since this concentrations of CO2, CH4, N2O, CFC-11 and CFC-12 model (and most other GCMs without trace gases) show a would have increased by 52, 89, 19, 49 and 121%, 110 Climate Modelling, Climate Prediction & Model Validation B
90N
60N h
30Nh
40 60 100 Time (years) 90N Figure B8: Cumulative spatial variance explained as a function of time and number of EOFs (1, 2, 10, 20 and 50) for Scenario "A" (see text for explanation). The signal is defined as the difference relative to the smoothed initial state (average over 60N h years 1-10) of the control run of the MPI model. Results are for annually averaged 2m temperature. (From Cubasch etai, 1991.)
(b)
150W 120W 90W 60W / / / /
I / V •iF \--s= 1 1 \ \ \''' \M , \/ \ - \ 1 \ 7 1 Figure 89: Distribution of the change in JJA soil moisture (cm) \ \ / / >' \ / 1 \*— 'y \\ simulated during successive 20-year periods over North America relative to a control by the GFDL coupled model vi'ith a 1 % per
- year (compound) increase of CO2: (a) years 40 to 60; (b) years 60
1 1.. 1 1 1 1 to 80, and (c) years 80 to 100. (From Manabe et al, 1992.) 60 50 40 30 20 10 0 • 10 20 30 40 Latitude
Figure BIO: The zonal mean of the surface fluxes of (a) heat, and (b) precipitation minus evaporation, simulated by the atmosphere in the UKMO coupled GCM (solid line) and the flux corrections (dashed line) added during the course of the coupled model integration. (From Murphy, 1990.) a,— a, pa
9, 9> 2 R R S <-J 5 Of) c o CO 00 ^ CO (N
S •3
00 00 10 O ON m r-- ^ ^ ir> —^ II 2 5 00 c O o 00 o ^ $iz S8 ii 6 ^' a, ON
0,00 X3 o 1^ T3 §-8 1 o •X II U O ^' •s — .5 E 'Co U o "S O il •i s o o s c o E 0 " & c < < < < < < o u u U u U 2 ? O I S u s 2 5 5^ ^ a o o i." XI 1 ^ c 11? 1 o c o (N o >o o ON ON in -I J J J 3 o IT) in in 0 2 CM X .11 in (u is x; a; 5 5 "I H ON o o u ON 8^ ON 8^ < --^ c 8^ _u > U . o r3 '3 cd a- O Figure Bll: Effect on equilibrium surface temperature change (DJF) using actual trace gases (Scenario "A", 1990-2050) instead of radiatively equivalent increases in CO2 (Wang et al, 1992): (a) warming using actual trace gases. Contour interval is 1 °C except in stippled region (>8°C) where the interval is 4°C; (b) difference due to use of actual trace gases rather than effective COj- Contour interval is 0.5°C, negative contours are dashed and areas of statistical significance are stippled. respectively, relative to 1990), the equilibrium simulations GCMs remains low, progress in the simulation of regional show nearly identical annual and global mean surface climate is being made with both statistical and one-way warming and increased precipitation (see Table B2). nested model techniques. In both cases the quality of the However, die regional pattern of climate changes between large-scale flow provided by the GCM is critical the two experiments is different (Figure Bll). Although The horizontal resolution in current general circulation the physical mechanisms for these differences have not yet models is too coarse to provide the regional scale been established, such results suggest that atmospheric information required by many users of climate change trace gases other than CO2 should be included explicitly in simulations. The development of useful estimates of future GCM simulations of both the present and future regional scale changes is dependent upon the reliability of climates. GCM simulations on the large-scale. Recognizing that simple interpolation of coarse-grid GCM data to a finer B2.4 Regional Climate Simulations grid is inadequate (Grotch and MacCracken, 1991), Although confidence in the regional changes simulated by strategies have been developed that involve the projection B Climate Modelling, Climate Prediction & Model Validation 113 of large-scale information from GCMs onto the regional scale either by using limited area models with boundary conditions obtained directly from the GCM (the one-way nested approach) or by using empirically-derived relationships between regional climate and the large-scale flow (the statistical approach). In both approaches, however, accuracy is limited by the accuracy of the large- scale flow generated in the GCM, and so they do not avoid the need for improvement in GCM simulations. Since the preparation of the IPCC (1990) report there has been considerable research using statistical approaches 1980 2000 2020 2040 2060 2080 Year to climate simulation; see, for example, Wigley et al. (1990), Karl et al. (1990) and Robock et al. (1991) Figure B12: Projected 5-year mean changes in precipitation amongst others. The statistical technique of Von Storch et over the Iberian peninsula (mm/month) in a transient CO2 al. (1991) has successfully reproduced observed rainfall experiment. Solid line = grid point values from a GCM; dashed patterns over the Iberian peninsula, and has been applied to lines = values from a statistical approach with large-scale patterns an IPCC Scenario A model integration (Cubasch et al, derived from the GCM. (From Von Storch et al, 1992.) 1991). The resulting changes in the winter rainfall are shown in Figure B12 as compared to the original GCM Figure B13: The average January surface air temperature (°C) over Europe: (a) as given by high-resolution observations; (b) as simulated by a mesoscale model nested within the NCAR CCMl R15; (c) as given by large-scale observations, and (d) as simulated by the NCAR CCMl alone. (From Giorgi et al, 1990.) 114 Climate Modelling, Climate Prediction & Model Validation B grid-point rainfall amounts; here, however, only about half (a) of the year-to-year variability is explained. Similar research is underway at other institutions (UKMO, BMRC). A recent review of research in these areas has been prepared by Giorgi and Meams (1991). The one-way nested modelling approach has been applied by Giorgi et al. (1990), whose regional scale climate simulations shown in Figure 813 compare well with high-resoludon observations, and show a significant amount of detailed information that was not portrayed in the original GCM. (The limited area model is likely to produce significant improvements only in regions where topographic effects are substantial.) Similar results have been found by McGregor and Walsh (1991) with the CSIRO model and by other groups (MPI, UKMO). The application of the nesting approach to regional climate changes in increased CO2 experiments is not quite so advanced, although Giorgi et al. (1991) have recently used the NCAR doubled CO2 simulation of Washington and Meehl (1991) in this way; their simulated changes of January surface air temperature for both the original GCM and the GCM plus a one-way nested mesoscale model are shown in Figure B14. In spite of this progress, the relatively low confidence attached to regional projections of any sort should be emphasized. Analysis by Karl et a/. (1991) has shown that the observed changes of temperature and the winter-to- summer precipitation ratio over central North America are not consistent with the IPCC (1990) projections, though 10°W 0° 10°E 20°E this inconsistency could be due to factors other than model errors. Figure B14: The distribution of changes in January surface air temperature over Europe as a result of doubled CO2: (a) the B3 Advances in the Analysis of Climate Feedbaclcs and temperature change (°C) given by the NCAR GCM relative to the Sensitivity GCM control; (b) the temperature change (°C) given by the B3.1 Introduction nested mesoscale model relative to the nested model control. While climate simulation continues apace, new attention (From Giorgi era/., 1991.) has been given to the critical role of feedback processes in determining the climate's response to perturbations. In general, water vapour is expected to amplify global Ramanathan (1989) using satellite data from regions of warming, while the effect of clouds remains uncertain. clear skies. The theoretical maximum concentration of Overall, there is no compelling evidence to change earlier water vapour is governed by the Clausius-Clapeyron model estimates of the climate's sensitivity to increased relation, and increases rapidly with temperature (about greenhouse gases. 6%/°C); this is the physical basis for the strong positive water vapour feedback seen in present climate models B3.2 Water Vapour Feedback (whereby increases in temperature produce increases in There is no compelling evidence that water vapour atmospheric water vapour which in turn enhance the feedback is anything other than the positive feedback it has greenhouse effect leading to a warmer climate). Increases generally been considered to be, although there may be in water vapour will be more effective at higher levels difficulties with the treatments of upper-level water vapour where temperatures are lower, and the change in emissivity in current models. (or effectiveness in absorbing thermal radiation) will be The importance of water vapour feedback in climate approximately proportional to the fractional change in change has long been recognized (Manabe and Wetherald, water vapour. (This simple relationship is complicated by 1967) and the dependence of the current greenhouse effect absorption by the water vapour continuum in the tropics, on water vapour has been documented by Raval and so that for a given percentage increase in water vapour, the B Climate Modelling, Climate Prediction & Model Validation 115 Figure B15: Vertical profiles of the global mean change in (a) specific humidity, and (b) relative humidity, simulated by the GISS GCM in a warm climate (+2°C SST change) relative to a cold climate (-2°C SST change). The full line denotes the standard version of the model and the dashed line denotes an improved version. (From Del Genio etal, 1991.) biggest contribution to increasing the greenhouse effect water vapour feedback (Cess et al., 1990) with relative occurs from about 500 hPa in cold regions, but from about humidity remaining constant to a first approximation 800 hPa in the tropics (Shine and Sinha, 1991).) (Mitchell and Ingram, 1992). Lindzen (1990) noted, The main source of water vapour is evaporation over the however, that at warmer temperatures, tropical convective oceans, and humidities in the oceanic boundary layer are clouds might detrain moisture at higher, colder close to saturation, with the concentration decreasing temperatures, and thus provide less water to the rapidly with height. Both convective and large-scale atmosphere directly. Del Genio et al. (1991) diagnosed motions and the evaporation of precipitation help to results from two versions of the GISS GCM, and found maintain the moisture distribution of the middle that in a warmer climate, cumulus-induced subsidence troposphere (Betts, 1990; Pierrehumbert, 1991; Kelly et indeed tends to increase drying as anticipated by Lindzen, al, 1991). The generally strong vertical gradients of water but found that this effect is offset by increased moistening vapour may, however, lead to systematic errors in models as a result of upward moisture transport by large-scale with parametrized vertical moisture diffusion or coarse eddies and by the tropical mean meridional circulation vertical resolution. Consequently, the large-scale transport (Figure B15). of moisture must be treated carefully in GCMs lest Estimates of the strength of the water vapour feedback spurious sources and sinks of moisture occur (Rasch and can be made from observations, although they are not Williamson, 1990, 1991; Kiehl and Williamson, 1991). independent of the effects of atmospheric motions. As Nevertheless, the overall realism of the outgoing clear-sky noted earlier, Raval and Ramanathan (1989) computed the radiative fluxes and of the simulated mean tropospheric dependence of the normalized clear-sky greenhouse effect temperatures suggests that the models' simulated derived from ERBE satellite data against sea surface tropospheric water vapour concentrations are not greatly in temperature from different locations and seasons. The error (Cess et al., 1990; Boer et al., 1991a). enhancement due to ascent in the intertropical convergence Under global warming, there is general agreement that zone (at temperatures around 303K) and the reduction due the atmospheric boundary layer would become moister, to subsidence (at temperatures around 295K) can be seen and as a result of both large-scale and convective mixing, in Figure B16. The overall positive slope suggests that the the humidity in the upper troposphere is also likely to greenhouse effect indeed increases with temperature, increase. All current GCMs simulate a strong positive although work by Arking (1991) indicates that some of the B Climate Modelling, Climate Prediction & Model Validation 117 maritime clouds. If tiiis correlation holds for global cycle does not strengthen as much as in other models warming, then it represents a positive feedback mechanism (Boer, 1991). associated with low cloudiness in these regions. (Note, however, that other processes can lead to local changes in B3.4 Surface Albedo Feedback cloud optical depth, and the correlation with temperature The conventional explanation of the amplification of may be fortuitous.) Feigelson's data, however, suggest global warming by snow feedback is that a warmer climate little change in optical depth with increasing temperature will have less snow cover, resulting in a darker surface for water clouds. which in turn absorbs more solar radiation. A recent Ramanathan and Collins (1991) have discussed a analysis suggests that snow-albedo-temperature feedback negative feedback mechanism associated with the processes in models are somewhat more complex than this production of highly reflective cirrus clouds as a view would indicate. consequence of an increase in tropical convective activity Using a methodology developed for an earlier study of associated with the 1987 El Niiio. The importance of this GCM cloud feedbacks (Cess et al., 1990), a perpetual mechanism for the case of greenhouse warming remains to April simulation was used with an imposed globally- be fully investigated, as the circulation changes in the uniform SST perturbation of 2°C relative to a prescribed tropics accompanying global warming are unlikely to be climatological SST distribution. To isolate snow feedback, the same as those during an El Niiio (Mitchell, 1991; Boer, two such April climate change simulations were 1991). performed, one in which the snow cover was held fixed Cloud feedback effects have been investigated in GCM and one in which the snow line was allowed to retreat studies in order to understand some of the consequences of following the imposition of the SST anomaly. The results the different ways in which clouds and cloud optical of these simulations are shown in Figure B18 in terms of properties are represented in models. Clouds are the climate feedback or sensitivity parameter X that was represented either diagnostically as a function of relative defined in Section 3.3.1 of the IPCC (1990) report. (An humidity, or prognostically by carrying an equation for increase in \ represents an increased climate change cloud water. Additionally, the optical properties of the caused by a given climate forcing.) There are clear clouds may be specified to remain fixed or they inay be differences among current GCMs in their response to parametrized to change as the climate changes (see Table changes in snow cover as measured by the ratio XIX^ Bl; also Table 3.2a, in IPCC 1990). As noted in IPCC (where \ is the sensitivity parameter for fixed snow). Here 1990, Cess et al. (1990) included examples of each of a value of 1,11.^ >1 indicating a positive snow feedback is these possibilities in a study of 19 GCMs, and found that found in most models. Also shown are the corresponding the simulated cloud feedback varied from slightly negative feedbacks in the case of an equivalent cloudless to strongly positive. atmosphere, found by separately averaging the models' Cloud feedback due to changes in cloud amount and/or clear-sky radiative fluxes at the top of the atmosphere distribution may be different, depending on whether clouds (TOA) in order to isolate the effects of clouds on snow are specified in terms of relative humidity or derived from an explicit cloud water prognostic equation. Differences may also arise from the differing treatment of clouds in explicit cloud water schemes. For example, in the UKMO 2.2 I 1 model (Senior and Mitchell, 1992a) increasing the lifetime 2.0 - • Global O of water cloud relative to ice cloud weakens the cloud O Clear • 1.8 - feedback (tending to reduce climate sensitivity), while in « 1.6 - • O the LMD model (Li and Le Treut, 1991) lowering the temperature at which ice cloud forms strengthens the cloud ^ 1,4 - O ^ • feedback. In the UKMO model the parametrization of cloud properties (as a function of cloud water content) appareittly results in a modest positive contribution to the 1.0 ^ total feedback in the tropics due to increased infrared 0.81 I I i_j 1 1 1 1 1 1 I I—I I I 1 1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 longwave emissivities. In the CCC model, on the other Model number hand, the parametrization of cloud properties (as a function of temperature) gives a negative contribution to the Figure B18: The snow feedback in terms of the ratio of the feedback in tropical regions due to the increase in cloud sensitivity parameter ("k) to that with fixed snow (A,j), as albedo. The reduced tropical solar input at the surface in simulated in 17 atmospheric GCMs. The full circles denote the perturbed climate thus moderates both the warming global values while the open circles are for cleai-sky conditions. and the increase in evaporation, so that the hydrological (From Cess era/., 1991.) 775 Climate Modelling, Climate Prediction & Model Validation B feedback. The clear-sky values of XIX^ range from of feedbacks that act to enhance or suppress the radiative negligible snow feedback to as much as a two-fold warming. The IPCC "best guess" for the climate sensitivity feedback amplification. is 2.5°C, with a range of uncertainty from 1.5 to 4.5°C. In general, clouds are thought to reduce positive snow Values of the climate sensitivity estimated from recent feedback by shielding the TOA albedo change. This equilibrium GCM simulations for doubled CO2 are mechanism, however, cannot account for the feedback sign summarized in Table 82, and generally fall within the reversals or the cloud-induced amplification of positive range given in Table 3.2a of IPCC 1990. These recent snow feedback exhibited by some models. As discussed by simulations convey little new information as they all (with Cess et al. (1991), the sign reversal in some cases is caused the exception of the LMD model) employ a relative by cloud redistribution, while in others it occurs as a humidity-based cloud scheme and fixed cloud radiative consequence of cloud-induced longwave interactions; there properties as in earlier models. The coupled model results is also a significant longwave feedback associated with (at the time of COT doubling) give lower bounds, as snow retreat in some models. These results are discussed in Section B2.2. complementary to those of Cohen and Rind (1991), who Recent additional estimates of the climate sensitivity found that the suppression of surface air temperature have been made by fitting the observed temperature record changes by positive snow anomalies is considerably to the evolution of temperature produced by simple reduced by a negative feedback involving the non-radiative energy-balance climate/upwelling-diffusion ocean models, components of the surface energy budget. assuming that all the observed warming over the last Uncertainties in the size of the surface albedo feedback century or so was due solely to increases in greenhouse associated with sea-ice remain. As Ingram et al. (1989) gases (see IPCC (1990), Sections 6 and 8). Schlesinger et noted, estimates of the strength of feedback depend not al. (1991) estimate a climate sensitivity of 2.2+0.8°C only on the model used, but also on the method used to allowing for the effect of sulphate aerosols, while Raper et provide the estimate. Covey et al. (1991) estimated an al. (1991) obtain a value of 2.3°C with a larger range of upper limit to sea-ice-albedo feedback by carrying out uncertainty due to their allowance for natural variability simulations in which the albedo of sea-ice was changed to (see Wigley and Raper, 1990) and they estimate a value of that of the open ocean. They found an enhancement of 1.4°C with no aerosol effect. The effect of including globally and annually averaged absorption of radiation of 2 sulphates (or other factors that could act to oppose the to 3Wm"2, comparable to the change on doubling CO2. greenhouse warming) in such calculations is to increase They concluded that for a warming of the magnitude the estimated climate sensitivity, since the observed expected on doubling CO2, sea-ice-albedo feedback is climate warming is then ascribed to a reduced radiative likely to be smaller than the feedback from water vapour forcing (see Sections A2.6 and C4.2.4). and potentially smaller than that from clouds. This is In summary, there have been a number of further studies consistent with Ingram et al. (1989) who estimated a that have a bearing on estimates of climate sensitivity. feedback of 0.2-0.3Wm"2K"l from sea-ice-albedo changes, New equilibrium GCM simulations have widened the whereas estimates of the strength of water vapour feedback range slightly to 1.7°C (Wang et al, 1991a) and 5.4°C are typically about l.SWm-K"' (see for example, (Senior and Mitchell, 1992a), but no dramatically new Mitchell, 1989). Meehl and Washington (1990) found that sensitivity has been found. Energy-balance model changing their sea-ice albedo formulation to one which considerations bring previous estimates of sensitivity was liable to induce ice melt led to both a warmer control (IPCC. 1990) more in line with the IPCC "best guess". climate (about IK) and a greater sensitivity (about 0.5K) to doubling CO2. Further discussion of sea-ice modelling is B4 Advances in Modelling Atmospheric Variability given in Section B5.4. 1. Although studies are incomplete in many respects, the ability of models to replicate observed atmospheric B3.5 Climate Sensitivity behaviour on a wide range of space and time-scales is There is no compelling new evidence to warrant changing encouraging. As noted in IPCC 1990, this ability provides the equilibrium sensitivity to doubled CO2 from the range some evidence that models may be usefully applied to of 1.5 to 4.5T as given by IPCC 1990. problems of climatic change associated with the The climate sensitivity is defined as the equilibrium greenhouse effect. change in global average surface air temperature due to a doubting of CO2 (Section 3.2 of IPCC, 1990), and is a 2. Model experiments with doubled CO2 give no clear measure of the response of a climate model to a change in indication of a systematic change in the variability of radiative forcing. The climate sensitivity may be thought temperature on daily to interannual time-scales, while the of as partly a direct radiative effect (estimated to be of the changes of variability for other climate features appear to order of 1.2°C for a doubling of CO2) and partly the effect be regionally (and possibly model) dependent. B Climate Modelling, Climate Prediction & Model Validation 119 B4.1 Introduction On doubling CO2, Cao et al. (1992) found a general Since tiie 1990 IPCC report there has been increased reduction in day-to-day temperature variability in winter in emphasis of the analysis of the ability of models to mid- to high latitudes over North America and over the simulate atmospheric variability on time-scales ranging ocean. The reduction was particularly pronounced in from diurnal to decadal in both control and perturbed regions where winter sea-ice had melted. Mearns et al. climate simulations. Such studies are critical to the (1990) , however, found no clear pattern of change in daily assessment of the effects of climate change, since it is temperature variability whereas P. Whetton (personal increasingly recognized that changes in variability and the communication) found general increases in daily occurrence of extreme events may have a greater impact temperature variability throughout Australia in both winter than changes in the mean climate itself (Katz and Brown, and summer. Mearns et al (1990) found a general increase 1991). Notwithstanding these results, there is at present in the day-to-day variability of precipitation due to only limited confidence in the ability of climate models to increased CO2, whereas Boer et al. (1991b) found an infer changes in the occurrence of interannual and regional overall decline in the daily variance of sea-level pressure events. due to doubling CO2, with the largest decrease occuiring The following discussion is arranged in order of in the North Atlantic in winter. increasing time-scales from diurnal to decadal. Particular attention is given to new studies that include simulated B4.4 Extreme Events changes in variability due to doubling atmospheric CO2, The simulation of extreme events is an important aspect of and to relating new results to the findings of IPCC 1990. a model's performance, and is closely connected with the Other studies are included to document the improvement question of natural variability. An important extreme event in the capability of models to simulate selected aspects of is the occurrence of tropical cyclones (or hurricanes) atmospheric variability. which GCMs cannot simulate in detail, though they do simulate tropical disturbances (see IPCC 1990, Section B4.2 Diurnal Cycle 5.3.3). Broccoli and Manabe (1990) and Haarsma et al Inclusion of the diurnal cycle (now a feature of many (1992) found that the spatial and temporal distributions of GCMs) permits the simulation of high-frequency temporal modelled tropical disturbances are similar to those variations. For example, a recent simulation by Randall et observed. On doubling CO2 Haarsma et al (1992) found al (1991a) has demonstrated that the diurnal variation of that the number of simulated tropical disturbances the hydrologic cycle in the tropics can be realistically increased, with little change in their average structure and simulated, while Cao et al. (1992) have found that the intensity. However, as noted in IPCC 1990, Broccoli and simulated diurnal range of surface temperature is generally Manabe (1990) found an increase in the number of tropical comparable to observations except for a slight storms if cloud cover was prescribed, but a decrease if underestimate in the mid-latitudes. On doubling CO2, Cao cloud was generated within the model. et al. (1992) found a slight decrease in the globally- In the 1990 IPCC report a consistent increase in the averaged diurnal range of surface temperatures, as noted in frequency of convective precipitation at the expense of IPCC 1990, although locally they found that factors such large-scale precipitation was noted, with the implication of as reduced soil moisture, receding snow lines or reduced more intense local rain at the expense of gentler but more cloud cover could lead to an increase in the diurnal persistent rainfall events. This tendency has been found in temperature range with doubled C02. Historical recent simulations using the CSIRO model (Gordon et al, observations over 25% of the global land area show a 1991; Pittock et al, 1991) and a high-resolution UKMO decreased diurnal temperature range, although the reasons model (J. Gregory, personal communication). Pittock et al for this change, which is largely a result of an increase in (1991) find a systematic increase in the frequency of heavy minimum temperatures, are not yet clear (see Section rain events with doubled CO2 (Figure B19) and a C3.1.5). consequent decrease in the return period of heavy rainfall (Figure B20). These changes are related to a systematic B4.3 Day-to-Day Variability increase in the frequency of penetrating convection in the In several of the models used in CO2 studies, the day-to• tropics and mid-latitudes on doubling CO2 (see IPCC, day variability of surface temperature is greater than that 1990). observed, particularly in higher northern latitudes (Meehl and Washington, 1990; Portman et al, 1990). Cao et al. B4.5 Blocking and Storm Activity (1992) found that in a UKMO model the daily variability Studies of the simulated variation of storm activity with of surface air temperature is overestimated in high enhanced CO2 are difficult to generalize as they use northern latitudes in winter, but is less than observed over different measures for storminess and address different mid-latitude continents in summer. regions. R. Lambert (personal communication) found a Climate Modelling, Climate Prediction & Model Validation B 120 160-| Renwick (1990) conducted a detailed analysis of the • 40°N [O 40'"S g Australian land points 140- results of the CSIRO model in die Australian-New Zealand >, o 120- region, and found a slight increase in the number of storms c 0) 100- in this region for all seasons. a> 80- The simuladon of persistent large-scale anomalies such c 60- as blocking has proved difficult to forecast with extended- B4.6 Intra-Seasonal Variability An important intra-seasonal variabihty is that associated E with the 30-60 day oscillation. This oscillation appears to influence onset and break phases of the monsoons, and thus has a considerable impact on the prediction of tropical Ic rainfall (Lau and Peng, 1990). These oscillations may also play a role in the onset of El Nino events. Park et al. Q (1990) and Zeng et al. (1990) have shown that some GCMs simulate intra-seasonal variations that are similar in structure to those observed, but they are generally of smaller amplitude. V V V 2.1 mth 3.9 mth 7.2 mth 1.1 yr 2.5 yr B4.7 Interannual Variability Return period As in IPCC 1990, no meaningful change in the interannual variability of temperature has been found with increased Figure B20: Daily rainfall amount vs. return period in Australia CO2, apart from a general reduction in the vicinity of the as simulated by the CSIRO model for both control (IXCO2) and winter sea-ice margins. doubled COj. (From Pittock et al., 1991.) Some of the models used in climate studies exaggerate the interannual variability of surface temperature in high latitudes in winter (Meams, 1991, reporting on simulations significant reduction in the number of cyclonic events in by Oglesby and Saltzman, 1990; Cao et al, 1992). The the CCC model between 30°N and the North Pole during simulated changes in the variability of temperature with winter, but a significant increase in the number of strong doubled CO2 vary from model to model; Rind (1991) cyclones. (Boer et al. (1991b) noted a decrease in the found decreases over much of the northern continents in variance of 1000 hPa height in the same experiment; see winter in the GISS model, as did Georgi et a/. (1991) using Section B4.3.) Senior and Mitchell (1992b) found a a regional model driven by output from the NCAR GCM northwest shift in the filtered variance of geopotential simulations of Washington and Meehl (I99I). On the odier height (a surrogate for storm tracks) in the North AUantic hand, Meams (1991) found both increases and decreases in in winter in the high resolution UKMO model. B. Hoskins temperature variability, while Cao et al (1992) found that (personal communication), on the basis of experiments increases were widespread in the tropics and subtropics, using the time-mean data from Senior and Mitchell's and in summer over North America and westem Europe, simulations, attributed the changes in storm tracks to with significant reductions confined to regions where sea- changes in the horizontal temperature gradient and ice receded in winter. increases in atmospheric water vapour. MuUan and B Climate Modelling, Climate Prediction & Model Validation 121 4.5 -" r 1 r —1—1—1—1—1—1—in—1—1—1—1—1—\—1— _(a) 3.0 h GCM OBS A\ A h \ /Vi* AvA /! \ I \ i\ / ' ll M hP\i.^ 1 LA \ d -1.5h 'A A/ VI ' V -4.5 —1 1 1 i_ 1 1 L 1 1 1 1 1 L 1 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 Year J I L. 70 71 72 73 74 75 76 77 78 79 80 81 82 V 83 84 85 86 87 88 89 Year Figure B21: The 1970-1989 lime-series of (a) the zonal surface wind stress anomahes averaged between 180°-140°W and 4°N-4°S, and (b) the Southern Oscillation Index. The observed variation is given by the dashed line, and that simulated by the MRI GCM is given by the solid line. (From Kitoh, 1991.) B4.8 ENSO and Monsoons Using tropical ocean-atmosphere models, Lau et al. A dominant mode of atmospheric variability in the tropics (1991), Nagai et al (1991), Philander et al (1991) and is that associated with tropical sea surface temperature Latif et al. (1991) have successfully simulated interannual anomalies in the Pacific. These tropical ocean-global variations that resemble some aspects of ENSO atmosphere oscillations are called El Niiio-Southern phenomena, although the amplitude of the simulated SST Oscillation, or ENSO. As noted in IPCC (1990), anomalies is generally less than that observed (as noted in atmospheric models are capable of giving a realistic IPCC, 1990) and model resolution appears to have an simulation of the seasonal tropical atmospheric anomalies important influence on the simulated behaviour. Meehl at least for intense El Niiio periods if they are given a (1990) claims some success in simulating ENSO-like satisfactory estimate of the anomalous sea surface disturbances in a global low-resolution ocean-atmosphere temperature (SST) in the tropical Pacific. This conclusion GCM, though it should be recalled that not all tropical SST is supported by more recent results (Konig et al, 1990; variations are associated with ENSO. As shown in Figure Kitoh, 1991; B. Hunt, personal communication). For B22, the patterns of tropical sea-level pressure changes example, as illustrated in Figure B21, a rather good that are characteristic of ENSO in the present climate are simulation has been made of the observed interannual also present with doubled COj (Meehl et al, 1991b). The variations of zonal surface wind stress over the central patterns of the tropical precipitation and soil-moisture equatorial Pacific and of the observed large-scale anomalies are similar in the control and doubled CO2 interannual variations of sea-level pressure as represented cases, with the dry areas becoming generally drier and the by the Southern Oscillation Index. Using prescribed sea wet areas becoming generally wetter with increased CO2. surface temperatures from 1987 and 1988, the broad These results, however, are yet to be confirmed with other characteristics of monsoon interannual variability have models. On increasing the concentration of greenhouse been reproduced (Palmer et al, 1991; Kitoh, 1992). Using gases in the MPI global coupled model, Lai et al. (1992) the coupled lAP model with observed initial conditions, found no evidence for a significant change in the mean Zeng et al (1990) and Li et al. (1991) report a successful onset date of the Indian Monsoon or for changes in the simulation of summer precipitation anomalies in the precipitation in the monsoon region. monsoon region of southeast Asia. 722 Climate Modelling, Climate Prediction & Model Validation B (a) 0 90E 180 90W 0 (b) Figure B22: The change in DJF sea-level pressure (hPa) between composite warm ENSO events and the 15-year mean simulated by a coupled ocean-atmosphere model for (a) normal CO2, and (b) doubled CO2. (From Meehl et al, 1991b.) B4.9 Decadal Variability the control run of the GFDL coupled model that was used There are now a sufficient number of model integrations in the transient CO2 simulation discussed in Section B2.2. over 50-100 year (and longer) periods to provide In this integration the surface flux corrections applied at preliminary information on the simulation of atmospheric the ocean surface were evidentiy successful in preventing a decadal variability. (See Section B4.3 for a discussion of multi-decadal drift of the mean temperature. Such internal simulated decadal variability in the ocean.) Whether variability may mask at least part of the changes due to simple models, low-resolution GCMs coupled to an increased greenhouse gases. oceanic mixed layer (Houghton et al., 1991; B. Hunt, personal communication) or full global ocean-atmosphere GCMs (Section B2.2) are used, the presence of B5 Advances in Modelling the Oceans and Sea-Ice considerable natural variability (i.e., in the absence of B5.1 Introduction changes in external forcing) on decadal time-scales is Although there is generally less experience in ocean characterisdcally found in model control runs. Typical modelling than there is in atmospheric modelling, since the magnitudes of such decadal variations are several tenths °C preparation of the IPCC (1990) report there have been a in surface air temperature, as illustrated in Figure B2 for number of studies with high-resolution and other ocean B Climate Modelling, Climate Prediction & Model Validation 123 Figure B23: (a) The instantaneous 160m temperature at TC intervals, and (b) the volume transport streamlines at 10 Sv intervals, as simulated in an eddy-resolving global ocean GCM with 1/2° resolution. (From Semtner and Chervin, 1988.) models that have potentially important consequences for energy that agrees reasonably well with the GEOSAT the simulation (and identification) of climate change with altimetric observations, although the simulated amplitudes coupled ocean-atmosphere GCMs (Anderson and are generally too small. A qualitatively similar agreement Willebrand, 1991). of simulated and observed variability was found by Semtner and Chervin (1988) with a 1/2° model of the B5.2 Eddy-Resolving Models global ocean, examples of which are shown in Figure B23. Recent simulations with the WOCE community ocean First results from the FRAM 1/4° model for the Antarctic model (Bryan and Holland, 1989) for the North Atlantic Circumpolar Current (Webb et al, 1991) indicate that the with a 1/3° resolution have yielded a distribution of eddy distribution of maximum variability associated with 124 Climate Modelling, Climate Prediction & Model Validation B mesoscale eddies is reasonably well simulated, although CFCs. Using a model with flux adjustments, Manabe et al. the eddy minima appear to be too low. The FRAM model (1991) find that the effect of enhanced greenhouse also shows a strong meridional (Deacon) cell in the warming is to reduce the rate of deep water formation. Antarctic Ocean, with a transport of over 20Sv. The (Washington and Meehl (1989) also note a reduction in upwelling branch of this cell, which is seen in both high- deep water formation in a model without flux and low-resolution models, occurs at those latitudes where adjustments.) This in turn decreases the flow of warm the coupled ocean-atmosphere models show delayed surface water from the soudi and probably contributes to a greenhouse gas warming (see Section B2.2). A recent delay in the greenhouse response in the northern North experiment by Boning and Budich (1991) has shown that Atlantic noted earlier in Figure B4. the simulated eddy energy increases substantially (along As shown with ocean-only models, variability on with significant variability on time-scales of several years) decadal and longer time-scales is associated with the when the horizontal resolution is increased to 1/6°. This advection of salinity anomalies through the deep raises the question (of practical importance for climate convection regions (Marotzke and Willebrand, 1991; modelling) whether or not mesoscale ocean eddies must be Weaver and Sarachik, 1991; Weaver et al., 1991; resolved in climate calculations. Mesoscale effects are so Mikolajewicz and Maier-Reimer, 1990). In the GFDL inhomogeneous and so tenuously connected to the larger- coupled model (R. Stouffer, personal communication) scale currents that a simple parametrization of their effects similar variability on time-scales of order 50 years is seems unlikely. However, studies using models that omit found. In an integration started from different initial salinity variations suggest that the total meridional heat conditions, a GFDL model (Manabe and Stouffer, 1988) in transport is only slightiy influenced by eddies, since the fact converged to a second equilibrium solution in which eddy flux tends to be compensated by a modification of the the North Atiantic was both fresher and colder (as shown mean flow induced by the eddies themselves (Bryan, in Figure B24) with a thermohaline circulation that was 1991). Further research on the climatic (as opposed to the weakly reversed. The existence of multiple equilibrium synoptic) role of ocean eddies is clearly required. states for the global ocean circulation raises the question of their stability, i.e., how easily the system may change from B5.3 Thermohaline Circulation one state to another. Maier-Reimer and Mikolajewicz A great deal of interest has recently focussed on the (1989) used the uncoupled MPI ocean GCM to investigate behaviour of the thermohaline circulation of the North the Younger-Dryas event and found that the North Atlantic Atlantic, whereby North Atlantic deep water is formed. thermohaline circulation was sensitive to relatively small This sinking is part of a global system of ocean transports variations in the strength and location of the surface fresh known as the Conveyor Belt Circulation (Gordon, 1986). water input, such that a breakdown of the circulation could Many aspects of the thermohaline circulation have been occur within a few decades. Other recent experiments with simulated in a 1/2° global ocean model (Semtner and zonally averaged ocean models have demonstrated that Chervin, 1991), and its basic dynamical structure and transitions between multiple equilibrium states can be sensitivity are being illuminated by both theoretical studies triggered by relatively modest changes in the large-scale and simulations with idealized models. In the present precipitation (Marotzke and Willebrand, 1991; Stocker and climate, high-latitude winter cooling of relatively saline Wright, 1991). surface water in the North Atlantic causes it to sink and The possibility of the natural collapse of the North subsequently to move out of the basin by deep southward Atlantic thermohaline circulation has obvious implications currents; this flow is compensated by the northward flow for our ability to predict future variations of the climate of warm surface water from the tropics, whose salinity system on decadal and longer time-scales. The North may be increased relative to the other high-latitude oceans Atlantic circulation, however, appears to be more robust to by enhanced evaporation (Stocker and Wright, 1991). change than the ocean-only model results suggest. Atlantic Coupled ocean-atmosphere models produce sinking in and Pacific deep sea sediment cores show that, with the high latitudes in the North Atlantic, though it is not clear exception of a possible short interruption during the how realistically they do so. It may also be noted that Younger-Dryas event, no major break in the thermohaline Dixon et al (1991) have recently succeeded in simulating circulation has occurred since the end of the last ice age in the observed spreading of CFCs in the Southern Ocean spite of the variability that has occurred in both the with the same GFDL coupled model that was used in the atmosphere and ocean (see, for example, Keigwin et al, transient CO2 experiments; this increases confidence that 1991). Preliminary results from a long integration of a the vertical mixing parametrization in the oceanic part of coupled ocean-atmosphere model (R. Stouffer, personal the model is a reasonable approximation. However, it communication) suggest that the thermohaline circulation should be noted that heat may affect the vertical stability in is more stable in a coupled model than in an ocean-only the ocean and hence may not behave in the same way as model with restored SSTs. B Climate Modelling, Climate Prediction & Model Validation 125 120E 180 120W 60W 0 60E 120E Figure B24: The difference in (a) mean surface temperature (°C), and (b) surface salinity (parts per thousand), in the North Atlantic in an equilibrium solution of the GFDL coupled ocean-atmosphere model with an active thermohaline circulation, relative to an alternative solution without thermohaline circulation. (From Manabe and Stouffer, 1988.) B5.4 Sea-Ice Models models have shown that such models are generally less Ocean-only and coupled ocean-atmosphere models sensitive to changes in the thermal forcing than are the generally use thermodynamic sea-ice models with (at simplified sea-ice models used so far in climate studies most) only simple advection schemes for the sea-ice. (Hibler, 1984; Lemke et al., 1990). This reduced Experiments with sea-ice models coupled to mixed-layer sensitivity is a result of a negative feedback between the pycnocline models (Owens and Lemke, 1990) have shown sea-ice dynamics and thermodynamics. It is therefore that the net freezing rate may depend on the continuum- possible that the pronounced response of the polar regions mechanical properties used for the sea-ice. Inclusion of found in experiments with increased CO2 (see Section sea-ice dynamics may therefore be required to properly B2.2 and IPCC (1990), Section 5) will be modified by the model the surface boundary conditions of the ocean. A inclusion of ice dynamics. On the other hand, dynamic sea- model that contains most of the characteristics of sea-ice ice models are more sensitive than thermodynamic models believed to be relevant for climate investigations was to changes in the wind forcing, and most atmospheric proposed by Hibler (1979) and has recently been circulation models do not reproduce the observed wind reformulated in a simplified version by Plato and Hibler field in polar regions very well. Further improvement of (1990). sea-ice models therefore depends upon advances in the Experiments with dynamic-thermodynamic sea-ice parametrization of polar processes, including the effect of 126 Climate Modelling, Climate Prediction & Model Validation B fog and low stratus clouds on the polar radiation balance. It may also be noted that models typically filter their solutions in polar latitudes for purely numerical reasons, and this may tend to degrade the simulation of sea-ice. B6 Advances in Model Validation B6.1 Introduction 150°W120°W90°W60°W30°W 0° 30°E 60°E 90°E laCE 150°E Climate models show an increasing ability to simulate the current climate and its variations. This improvement is due to a combination of increased model resolution and improved physical parametrizations. The correctness of simulated results for the present climate at both global and regional scales is a desirable but not a sufficient condition for increased confidence in their use for simulations of future climate change. 150°W120°W90°W60''W30°W O" SCE SCE 90°E 120''E 150°E B6.2 Systematic Errors and Model Intercomparison (C) Every climate modelling group strives to identify their models' systematic errors as a natural part of the continuing process of model development. Ideally, each physical parametrization in a climate model should be individually calibrated against appropriate observational data, although in many (if not most) cases this is not 150''W120°W90''W60''W30°W 0° 30''E 60°E 90°E 120''E IBCE possible. Instead, with a particular set of parametrizations and a particular resolution, a model's performance is Figure B25: The 200 hPa velocity potendal anomalies (lO^m^s') evaluated against the available seasonal climatological for JJA 1988: (a) as given by ECMWF analyses referenced distributions of the large-scale variables such as against the mean during 1986-1990; (b) as simulated by the temperature and circulation. While atmospheric and ECMWF model over the same interval with observed SST and oceanic models continue to improve, the overall standard surface flux parametrizations for heat and moisture, and characterization of their performance given in Section 4 of (c) as simulated with revised parametrizations accounting for IPCC (1990) remains valid (Gates et al, 1990). The fluxes at low windspeeds. (From Miller et al, 1991.) situation is somewhat different, however, for coupled atmosphere-ocean GCMs, for which only a provisional error assessment is possible due to the limited number of simulated in the control runs of 14 atmospheric GCMs. integrations that have been performed (see below and Despite the fact that the models differed greatly in Section B2.2) and because of the limited ocean data resolution and in the nature and sophistication of their currently available. parametrizations of physical processes, a number of Progress in atmospheric model validation since the common systematic errors where found. The simulated IPCC (1990) report has occurred both in the form of climate was colder than that observed on average, and all documented improvements in specific GCMs and in the models showed a cold bias in the middle- and high-latitude form of intercomparisons of large numbers of independent upper troposphere and in the tropical lower stratosphere GCMs. An example of an improvement in a particular (see Figure B26). These temperature errors induce model is the change in the parametrization of tropical corresponding errors in the zonal wind distribution in evaporation and convection that has removed a major accordance with the thermal wind relationship, with the systematic error in the ECMWF model (see Figure 825). result that most GCMs display too westerly a flow in the Recent intercomparisons of atmospheric models have upper troposphere and in high latitudes. Taken as a whole, either focussed on the simulation of the present climate (as models appear to be more sensitive to changes in physical in Boer et al, 1991a) or have considered the radiative parametrizations than to modest changes in resolution. forcing induced by a prescribed climate change (as in Cess The principal features of the Hadley, Ferrel and polar et al, 1991) (see Section B3.3.3). On behalf of the cells that characterize the atmospheric circulation are Working Group on Numerical Experimentation (WGNE), simulated by all models, along with appropriate seasonal Boer et al. (1991a) have collected the mean DJF and JJA shifts. There is, however, considerable variation in the distributions of selected atmospheric variables as strengths of the models' simulated meridional circulations. B Climate Modelling, Climate Prediction & Model Validation 127 (a) JJA 1050 1 1 —1 1 I 1030 - 1010 S 1000 80N 60N 40N 20N 0 20S 40S 60S SOS Q. Latitude - 950 1 1 1 1 1 90N 80 30 0 30 60 908 Latitude (b) DJF SON SON 40N 20 N 0 20S 40S 608 SOS Latitude Figure B26: The systematic error (simulated minus observed) in mean temperature (°C) for (a) DJF, and (b) JJA for the UKMO atmospheric GCM. (From Boer et al, 1991a.) At the same time there has been improvement in the simulation of mean sea-level pressure in a number of models, due partly to increased resolution and partly to the introducdon of the parametrization of gravity wave drag. This improvement is illustrated in Figure B27a. The largest differences among models are found near Antarctica, where the extrapolation of pressure to sea level may introduce artificial variations. There has not been as much improvement in the simulation of precipitation by current GCMs. As shown in 90S Figure B27b, there is a systematic over-estimation of the precipitation over much of the Northern Hemisphere (as Figure B27: (a) The zonally-averaged JJA sea level pressure well as an apparent under-estimation in the mid-latitudes (mb) simulated by current high resolution GCMs; (b) the zonally- of the Southern Hemisphere), and the models' simulated averaged DJF precipitation (mm/day) simulated by current precipitation in the tropics varies considerably around that GCMs. Different symbols indicate different models and the solid observed. There are, however, uncertainties in the line is the observed climatological average according to Jaeger observed data themselves, especially over the tropical (1976). (From Boeref o/., 1991a.) oceans and in the Southem Hemisphere. Relatively large uncertainties in the simulation of regional precipitation are shown in the intercomparison of atmospheric GCMs being considerable variation is found in the models' simulated undertaken by the Monsoon Numerical Experimentation increases in precipitation, evaporation and total atmo• Group (MONEG) of the WCRP (1990b). spheric water vapour in response to a uniform 4°C SST In a companion study to the model intercomparisons of increase. As shown in Figure B28, the models also Cess et al. (1990, 1991) (see Section B3.3), an inter• simulate an overall increase in the net infrared clear-sky comparison of the surface energy fluxes in the GCMs has flux at the surface, reflecting the positive temperature/ recentiy been made (Randall et al., 1991b). In this study water vapour feedback (Section B3.2) in the models. 128 Climate Modelling, Climate Prediction & Model Validation B simulated rate and location of deep water formation and in the strength of deep ocean circulation. In order to examine ocean model performance in more detail, an intercomparison of tropical Pacific ocean models has recently been undertaken by TOGA-NEG (WCRP, 1990a; D. Anderson, personal communication), in which several models have been integrated with the same representation of surface forcing. In general, the SST error in all models shows a similar pattern, with the central equatorial Pacific being too cold and the equatorial western Pacific too warm throughout the year. Water along the western coast of South America is also generally too warm, and the equatorial undercurrent tends to be too -25 F I I I I I I I I I I I I I I I I I I I I I I I I I 1 I I I 90N 60 30 0 -30 -60 weak. These discrepancies may be at least partly caused by Latitude inadequate surface forcing. An intercomparison of the behaviour of the tropical Figure B28: The change in the zonally-averaged longwave Pacific SST in coupled ocean-atmosphere models (in clear-sky flux at the surface as simulated by 14 atmospheric which the ocean and/or the atmosphere is a GCM) has GCMs in response to a uniform 4°C increase of SST relative to recently been made by Neelin et al. (1992). These models July chmatology. (From RandaO et al, 1991b.) show a wide range of behaviour in their simulation of tropical interannual variability, and some are more successful in simulating ENSO-like events than others. In the Boer et al. (1991a) intercomparison, as in most The runs were not made under the same conditions, previous intercomparisons of climate models, results have however, and there is a wide range of model sophistication been collected as available from the modelling community represented. and no attempt has been made to run the models under The comparison of simulated past climates with palaeo common conditions and for common periods of time. In an data provides a further test of models, though the effort to achieve a more systematic intercomparison of usefulness of such tests is limited by the quality of data models, these attributes have recently been incorporated and by our knowledge of the changes of past climates. A into the WGNE/PCMDI Atmospheric Model Inter• Palaeoclimate Model Intercomparison Project (PMIP) has comparison Project (AMIP), in which virtually all of the been proposed (NATO Advanced Research Workshop on world's atmospheric GCMs will simulate the decade 1979- Palaeoclimatic Modelling, 27-31 May 1991, Saclay, 1988 with common values of the CO2 concentration and France) to promote the understanding of the response of solar constant and with observationally-prescribed but climate models to past changes in forcing. monthly-varying SST and sea-ice distributions (Gates, Although model intercomparison of the sort described 1991). The AMIP involves many of the same modelling here does not address many important questions in climate groups that participated in the earlier WCRP modelling (such as how to determine which parametriz• Intercomparison of Radiation Codes in Climate Models ations are "best" or how best to couple the atmosphere and (ICRCCM), the results of which have recently been ocean), intercomparison has proven useful as a collective published (Ellingson et al, 1991; Fouquart et al, 1991). exploration of model parameter space and as a benchmark The identification of systematic errors and the for the documentation of model improvement. It is intercomparison of ocean models are at a less advanced characteristic of model intercomparisons that no single stage than for atmospheric models. When forced with model is found to be superior to all other models in all "observed" surface temperatures, salinities and wind respects. It should also be recognized that the various stresses, ocean models have been moderately successful in intercomparisons are designed for different purposes: those reproducing the observed large-scale circulation and water of Cess et al. (1990, 1991) are concerned with the models' mass distribution (see Section B5). The principal response to changes in forcing (and are therefore relevant systematic errors common to most ocean models are an to the response to possible future changes of greenhouse underestimate of the meridional heat transport as inferred gases), while those of Boer et al (1991a) and those being from observations, and the simulation of the main undertaken in AMIP (Gates, 1991) are concerned with the thermocline to be too deep and too diffuse. No common models' ability to simulate the present climate. As model systematic errors have as yet been identified in the deeper development continues (and as increased computer ocean in view of the paucity of observational data, resources become available), there will be both an although there are considerable differences in the opportunity and a need to conduct further climate model B Climate Modelling, Climate Prediction & Model Validation 129 intercomparisons, especially of coupled ocean-atmosphere development and validation of global atmospheric and models in both their control configuration and in oceanic models, and have culminated in the proposal for a experiments with increasing greenhouse gases. Global Climate Observing System (GCOS) (WCRP, 1991). B6.3 Data for Model Validation The availability of appropriate observational data is a critical factor in the validation (and improvement) of climate models, and some progress has been made in the assembly of global data sets for selected climate variables. References The recent global compilations of average monthly surface air temperature and precipitation by Legates and Willmott Anderson, D.L.T. and J. Willebrand, 1991: Recent advances in (1990a, b) are believed to be improvements over earlier modelling the ocean circulation and its effects on climate. Rev. Progress Phys. (In press). atlases of these variables, and new assemblies of land- Arking, A., 1991: Temperature variability and water vapor based precipitation (Hulme, 1992) and soil moisture feedback. (Unpublished manuscript). (Vinnikov and Yeserkepova, 1991) are available. It should Betts, A.K. and Harshvardhan, 1987: Thermodynamic constraint also be noted that the diagnostics made from the on the cloud liquid water feedback in climate models. J. operational analyses of global numerical weather Geophys. Res., 92, 8483-8485. prediction models (Hoskins et al., 1989; Trenberth and Betts, A.K., 1990: Greenhouse warming and the tropical water Olson, 1988) are important sources of data for the budget. Bull. Amer Met. Soc, 71, 1464-1465. validation of selected aspects of atmospheric models. The Boer, G., 1991: Chmate change and the regulation of the surface climatological ocean atlas of Levitus (1982) has been moisture and energy budgets. (Unpublished manuscript). supplemented by the compilation of surface variables from Boer, G.J., K. Aipe, M. Blackburn, M. Deque, W.L. Gates, T.L. the COADS data set (Wnght, 1988; Oberhuber, 1988; Hart, H. Le Trent, E. Roeckner, D.A. Sheinin, I. Simmonds, Michaud and Lin, 1991). These data are proving useful in R.N.B. Smith, T. Tokioka, R.T. Wetherald and D. WiUiamson, 1991a: An intercomparison of the climates simulated by 14 the validation of both ocean and coupled ocean- atmospheric general circulation models. CAS/JSC Working atmosphere models, as are the observations of transient Group on Numerical Experimentation Report No. 15, tracers in the ocean (Toggweiler et al, 1989; Dixon et al, WMO/TD - No. 425, World Met. Organiz., Geneva, 37 pp. 1991). Boer, G.J., N.A. McFarlane and M. Lazare, 1991b: Greenhouse For other variables such as cloudiness, precipitation, gas induced climate change simulated with the Canadian evaporation, run-off, surface heat flux, surface stress, Climate Centre second generation general circulation model. J. ocean currents and sea-ice, however, the observational data Climate (In press). base remains inadequate for the purposes of model Boning, C.W. and R. Budich, 1991: Eddy dynamics in primitive validation. These variables are relatively difficult to equation models: Sensitivity to horizontal resolution and observe on a global basis, and are not easily inferred from friction. (Unpublished manuscript). compilations of conventional circulation statistics such as Broccoli, A.J. and S. Manabe, 1990: Can existing climate models those of Oort (1983). The best prospect for their systematic be used to study anthropogenic changes in tropical cyclone climate? Geophys. Res. Letters, 17, 1917-1920. global estimation is probably the procedure known as re- Bryan, F. and W.R. Holland, 1989: A high-resolution simulation analysis, whereby modern data assimilation techniques are of the wind- and thermohaline-driven circulation in the North used to retrospectively initialize a comprehensive Atlantic Ocean: In Parametrization of Small-scale Processes, atmospheric GCM on a daily basis for a number of past Proceedings of 'Aha Huliko'a Hawaiian Winter Workshop. years. Such projects are being planned by both ECMWF Univ. Hawaii, Manoa, pp. 17-20, 99-115. and NMC, and the possibility of applying the technique to Bryan, K., 1991: Poleward heat transport in the ocean: A review ocean models is under active consideration. of a hierarchy of models of increasing resolution. Tellus, 43 A- Finally, it should be noted that the development of a B, 104-115. more comprehensive global climate data base is a key Cao, H-X., J.F.B. Mitchell and J.R. Lavery, 1992: Simulated element of the World Climate Research Programme diurnal range and variability of surface temperature in a global (WCRP). The observational plans that have been climate model for present and doubled COo. J. Climate. (In press). developed in recent years for the Global Precipitation Cess, R.D., G.L. Potter, J.P. Blanchet, G.J. Boer, A.D. Del Climatology Project (WCRP, 1988a), for the World Ocean Genio, M. Deque, V. Dymnikov, V. Galin, W.L. Gates, S.J. Circulation Experiment WOCE (WCRP, 1988b), for the Ghan, J.T. Kiehl, A.A. Lacis, H. Le Trent, Z.-X. Li, X.-Z. Tropical Ocean-Global Atmosphere project TOGA Liang, B.J. McAvaney, V.P. Meleshko. J.F.B. Mitchell, J.-J. (WCRP, 1990a), and for the Global Energy and Water Morcrette, D.A. Randall, L. Rikus, E. Roeckner, J.F. Royer, U. Cycle Experiment GEWEX (WCRP, 1990c) are focussed Schlese, D.A. Sheinin, A. Slingo, A.P. Sokolov, K.E. Taylor, on the acquisition of data that are necessary for the further W.M. Washington, R.T. Wetherald, I. Yagai and M.-H. Zhang, 130 Climate Modelling, Climate Prediction & Model Validation B 1990: Intercomparison and interpretation of climate feedback regional climate simulation over Europe. J. Geophys. Res., 95, processes in nineteen atmospheric general circulation models. 18,413-18,431. J. Geophys. Res., 95, 16,601-16,615. Giorgi, F., M.R. Marinucci and G. Visconti, 1991: A 2xC02. Cess, R.D., G.L. Potter, M.-H. Zhang, J.-P. Blanchet, S. Chahta, climate change scenario over Europe generated using a R. Colman, D.A. Dazlich, A.D. Del Genio, V. Dymnikov, V. limited-area model nested in a general circulation model. Part Galin, D. Jerrett, E. Keup, A.A. Lacis, H. Le Treut, X.-Z. II. Climate change scenario. (Unpublished manuscript). Liang, J.-F. Mahfouf, B.J. McAvaney, V.P. Meleshko, J.F.B. Gordon, A.L., 1986: Inter-ocean exchange of thermocline water. Mitchell, J.-J. Morcrette, P.M. Norris, D.A. Randall, L. Rikus, J. Geophys. Res., 91, 5037-5046. E. Roeckner, J.-F. Royer, U. Schlese, D.A. Sheinin, J.M. Gordon, H.B., P. Whetton, A.B. Pittock, A. Fowler, M. Haylock Slingo, A.P. Sokolov, K.E. Taylor, W.M. Washington, R.T. and K. Hennessy, 1991: Simulated changes in daily rainfall Wetherald and I. Yagai, 1991: Interpretation of snow-climate intensity due to the enhanced greenhouse effect: ImpUcations feedback as produced by 17 general circulation models. for extreme rainfall events. (Unpublished manuscript). Science, 253, 888-892. Grotch, S.L. and M.C. MacCracken, 1991: The use of general Charlson, R.J., J. Langer, H. Rodhe, C.B. Leovy and S.G. circulation models to predict regional climate change. J. Warren, 1991: Perturbation of the Northern Hemisphere Climate, 4, 286-303. radiadve balance by backscattering from anthropogenic sulfate Haarsma, R.J., J.F.B. Mitchell and C.A. Senior, 1992: Tropical aerosols. Tellus, 43AB, 152-163. disturbances in a GCM. Climate Dyn. (In press). Cohen, J. and D. Rind, 1991: The effect of snow cover on the Hansen, A.R. and A. Sutera, 1990: Weather regimes in a general climate. J. Climate, 4, 689-706. circulation model. J. Atmos. Sci., 47, 380-392. Covey, C, K.E. Taylor and R.E. Dickinson, 1991: Upper limit Hansen, J.E., G. Russell, A. Lacis, I. Fung and D. Rind., 1985: for sea-ice albedo feedback contribution to global warming. J. Climate response times: Dependence on climate sensitivity and Geophys. Res., 96, 9169-9174. ocean mixing. Science, 229, 857-859. Cubasch, U., K. Hasselmann, H. Hock, E. Maier-Reimer, U. Hasselmann K., R. Sausen and E. Maier-Reimer, 1992: On the Mikolajewicz, B.D. Santer and R. Sausen, 1991: Time- cold-start problem in transient simulations with coupled dependent greenhouse warming computations with a coupledocean-atmosphere models. Max Planck Inst. Meteor. Report, ocean-atmosphere model. Max Planck Inst. Meteor. Report 67, Hamburg. (In press). Hamburg. Hibler, W.D., 1979: A dynamic-thermodynamic sea-ice model. Del Genio, A.D., A.A. Lacis and R.A. Ruedy, 1991: Simulations J. Phys. Oceanogr, 9, 815-846. of the effect of a warmer climate on atmospheric humidity. Hibler, W.D., 1984: Sensitivity of sea-ice models to ice and Nature, 251, 382-385. ocean dynamics. WCP-77, WMO, Geneva. Dix, M., B.J. McAvaney, I.G. Watterson and H.B. Gordon, 1991: Hoskins, B.J., H.H. Hsu, I.N. James, M. Masutani, P.D. Personnal communication. Full text in preparation. Sardeshmukh and G.H. White, 1989: Diagnostics of the global Dixon, K.W., J.L. Bullister, R.H. Gammon, R.J. Stouffer and atmospheric circulation based on ECMWF analyses 1979- G.P.J. Theile, 1991: Climate model simulation studies using 1989. WCRP-27, WMO/TD - No. 326, World Met. Organiz., chlorofluorocarbons as transient oceanic tracers. (Unpublished Geneva, 217 pp. manuscript). Houghton, D.D., R.G. Gallimore and L.M. Keller, 1991: Ellingson, R.G., J. ElUs and S. Eels, 1991: The intercomparison Stability and variability in a coupled ocean-atmosphere climate of radiation codes used in climate models: Longwave results. model; Results of 100-year simulations. J. Climate, 4, 557- J. Geophys. Res., 96, 8929-8954. 577. Feigelson, E.M., 1978: Preliminary radiation model of a cloudy Hulme, M., 1992; A 1951-80 global land precipitation atmosphere. Structure of clouds and solar radiation. Contrib. climatology for the evaluation of general circulation models. Atmos. Phys., 51, 203-229. Climate Dyn., 7, 57-72. Flato, G.M. and W.D. Hibler. 1990: On a simple sea-ice Ingram, W.J., C.A. Wilson and J.F.B. Mitchell, 1989: Modelling dynamics model for climate studies. Ann. GlacioL, 14, 72-77. climate change: an assessment of sea-ice and surface albedo Fouquart, Y., B. Bonnel and V. Ramaswamy, 1991: Inter- feedbacks. /. Geophys. Res., 94, 8609-8622. comparing shortwave radiation codes for climate studies. J. IPCC, 1990: Climate Change, The IPCC Scientific Assessment. Geophys. Res., 96, 8955-8969. J.T. Houghton, G.J. Jenkins and J.J. Ephraums (Eds.). Gates, W.L., 1991: The WGNE atmospheric model inter• Cambridge University Press. UK. 365pp. comparison project. AMIP Newsletter No. 1, PCMDI, Jaeger, L., 1976; Monatskarten des Niederschlags fur die ganze Lawrence Livermore Nat'l. Lab., Livermore, USA. 8pp. Erde. Ber. Deut. Wetterdienstes, Nr. 139, Offenbach. 38pp. Gates, W.L., P.R. Rowntree and Q.-C. Zeng, 1990: Validation of Karl, T.R., W.-C. Wang, M.E. Schlesinger, R.W. Knight and D. climate models. In: Climate Change: The IPCC Scientific Portman, 1990: A method of relating general circulation model Assessment. J.T. Houghton, G.J. Jenkins and J.J. Ephraums simulated climate to the observed local climate. Part I: (Eds.). Cambridge University Press, Cambridge. pp93-130. Seasonal statistics. /. Climate, 3, 1053-1079. Giorgi, F. and L.O. Mearns, 1991: Approaches to the simulation Karl, T.R., R.R. Heim and R.G. Quayle, 1991: The greenhouse of regional climate change: A review. J. Geophys. Res., 29, effect in central North America: If not now. when? Science, 191-216. 251, 1058-1061. Giorgi, F., M.R. Marinucci and G. Visconti, 1990: Use of a limited-area model nested in a general circulation model for B Climate Modelling, Climate Prediction & Model Validation 131 Katz, R.W. and B.G. Brown, 1991: Extreme events in a changing Li, Z-X. and H. Le Trent, 1991: Cloud radiation feedback in a climate: Variability is more important than averages. Climatic GCM and their dependence on cloud modelling assumptions. Change. (In press). Climate Dyn. (In press). Kelly, K.K., A.F. Tuck and T. Davies, 1991: Wintertime Lindzen, R.S., 1990: Some coolness concerning global warming. asymmetry of upper tropospheric water content between Bull. Amer. Met. Soc, 71, 288-299. Northern and Southern Hemispheres. Nature, 353, 244-247. Maier-Reimer, E. and U. Mikolajewicz, 1989: Experiments with Kiegwin, L.D., G.A. Jones, S.J. Lehman and E.A. Boyle, 1991: an OGCM on the cause of the Younger-Dryas. In: Deglacial meltwater discharge. North Atlantic deep circulation Oceanography 1988 A. Ayala-Castanares, W. Wooster and A. and abrupt climate change. J. Geophys. Res., 96, 16,811- Yanez-Arancibia (Eds.). UNAM Press, Mexico City, pp87-100. 16,826. Manabe, S. and R.J. Stouffer, 1988: Two stable equilibria of a Kiehl, J.T. and D.L. Williamson, 1991: Dependence of cloud coupled ocean-atmosphere model. / Climate, 1, 841-866. amount on horizontal resolution in the National Center for Manabe, S., M.J. Spelman and R.J. Stouffer, 1992: Transient Atmospheric Research Community Climate Model. J. responses of a coupled ocean-atmosphere model to gradual Geophys. Res., 96, 10955-10980. changes of atmospheric CO2. Part II: Seasonal response. Kitoh, A., 1989: Persistent anomalies in the Northern J Clim., 5, 105-126. Hemisphere of the MRI GCM-1 compared with the Manabe, S., R.J. Stouffer, M.J. Spelman and K. Bryan, 1991: observation. Pap. Met. Geophys., 40, 83-101. Transient responses of a coupled ocean-atmosphere model to Kitoh, A., 1991: Interannual variations in an atmospheric GCM gradual changes of atmospheric CO2. Part I: Annual mean forced by the 1970-1989 SST. Part I: Response of the tropical response. J. Climate, 4, 785-818. atmosphere. J. Met. Soc. Japan, 69, 251 -269. Manabe S. and R.T, Wetherald, 1967: Thermal equilibrium of Kitoh, A., 1992: Simulated interannual variations of the Indo- the atmosphere with a given distribution of relative humidity. Australian monsoons. J. Met. Soc. Japan, 70 (special edition J. Atmos. Sci, 2^, 241-259. on Asian Monsoon). (In press). Marotzke, J. and J. Willebrand, 1991: Multiple equilibria of the Konig, W., E. Kirk and R. Sausen, 1990: Sensitivity of an global thermohaline circulation. J. Phys. Oceanogr. (In press). atmospheric general circulation model to interannually varying McAvaney, B.J., R. Colman, J.F. Fraser and R.R. Dahni, 1991: SST. Ann. Geophys. Sci., 8, 829-844. The response of the BMRC AGCM to a doubling of C02- Lai, M., U. Cubasch and B.D. Santer, 1992: Potential impacts of BMRC Technical Memorandum No.3. (In press). global warming on monsoon climate as inferred from a McGregor, S.L. and K. Walsh, 1991: Summertime climate coupled ocean-atmosphere general circulation model. simulations for the Australian region using a nested model. (Unpublished manuscript). Proceedings Fifth Conf. Climate Variations, Amer. Met. Soc, Latif, M., M. Flugel and J-S. Xu, 1991: An investigation of short Denver, USA. range climate predictability in the tropical Pacific. Max Mearns, L.O., 1991: Changes in climate variability with global Planck Inst. Meteor., Report No. 52, Hamburg. warming and its possible impacts. Proceedings of the First Lau, K.M. and C. Peng, 1990: Origin of low-frequency (intra- Nordic Interdisciplinary Conference on the Greenhouse Effect, seasonal) oscillations in the tropical atmosphere. Part III, 16-19 September 1991, Copenhagen. (In press). Monsoon dynamics. J. Atmos. Sci., 47, 1443-1462. Mearns, L.O., S.H. Schneider, S.L. Thompson and L.R. Lau, N.-C, S.G.H. Philander and M.J. Nath, 1991: Simulation of McDaniel, 1990: Analysis of climate variability in general EI Niiio-Southern Oscillation phenomena with a low-resolution circulation models: Comparison with observations and changes coupled general circulation model of the global ocean and in variability in 2xC02 experiments. J. Geophys. Res., 95, atmosphere. (Unpublished manuscript). 20469-20490, Legates, D.R. and C.J. Willmott, 1990a: Mean seasonal and Meehl, G.A., 1990: Seasonal cycle forcing of El Niiio-Southern spatial variability in global surface air temperature. Theor. Oscillation in a global, coupled ocean-atmosphere GCM. J. Appl. Climatology, 41, 11-21. Climate, 3, 72-98. Legates, D.R. and C.J. Willmott, 1990b: Mean seasonal and Meehl, G.A. and W.M. Washington, 1990: CO2 climate spatial variability in gauge-corrected global precipitation. Int. sensitivity and snow-sea-ice albedo parametrization in an J Climatology, 10, 111-127. atmospheric GCM coupled to a mixed-layer ocean model. Lemke, P., W.B. Owens and W.D. Hibler, 1990: A coupled sea- Climatic Change, 16, 283-306. ice mixed-layer pycnocline model for the Weddell Sea. J. Meehl, G.A., W.M. Washington and T.R. Karl, 1991a: Low- Geophys. Res., 95, D6, 9513-9525. frequency variability and CO2 transient climate change. Part I. Le Trent, H., Z.X. Li, M. Foaichon and R. Butel, 1992: The Time-averaged differences. (Unpublished manuscript) sensitivity of the LMD GCM to greenhouse forcing: an Meehl, G.A., G.W. Branstator and W.M. Washington, 1991b: El analysis of the atmospheric feedback effects. (Manuscript in Nifio-Southern Oscillation and CO2 climate change. preparation) (Unpublished manuscript) Levitus, S., 1982: Climatological Atlas of the World Ocean. Michaud, R. and CA, Lin, 1991: Monthly summaries of NOAA Professional Paper 13, US Dept. Commerce, merchant ship surface marine observations and implications Washington, DC, USA. 173 pp. for climate variability studies. Climate Dyn. (In press) Li, X., Q.-C. Zeng and G.C. Yuan, 1991: Paper presented at Mikolajewicz, U. and E. Maier-Reimer, 1990: Internal secular Workshop of the Monsoon Numerical Experimentation Group, variability in an ocean general circulation model. Climate Boulder, USA. October 1991. Dyn., A, 145-156. 132 Climate Modelling, Climate Prediction & Model Validation B Miller, M.J., A.C.M. Beljaars and T.N. Palmer, 1991: The Pittock, A.B., A.M. Fowler and P.H. Whetton, 1991: Probable sensitivity of the ECMWF model to the parametrization of changes in rainfall regimes due to the enhanced greenhouse evaporation from the tropical oceans. / Climate. (In press) effect. Proceedings of Intn'l. Hydrology and Water Resources Mitchell, J.F.B., 1989: The "greenhouse effect" and climate Symp., Perth, Australia, October 1991. change. Rev. Geophys., 11, 115-139. Portman, D.A., W.-C. Wang and T.R. Karl, 1990: A comparison Mitchell, J.F.B., 1991: No limit to global warming? Nature, 353, of general circulation model and observed regional climates: 219-220. Daily and seasonal variability. Proceedings of 14th Climate Mitchell, J.F.B. and W.J. Ingram, 1992: On COj and climate: Diagnostics Workshop, NOAA, Washington, D.C, USA. pp Mechanisms of changes in cloud. J Climate., 5, 5-21. 282-288. Miyakoda, K. and J. Siruds, 1990: Subgrid scale physics in 1- Ramanathan, V. and W. Collins, 1991: Thermodynamic reg• month forecasts. Part II: Systematic error and blocking ulation of ocean warming by cirrus clouds deduced from the forecasts. Mow. Wea. Rev., 118, 1065-1081. 1987 El Nino. Nature, 351, 27-32. Mullan, A.B. and J.A. Renwick, 1990: Climate change in the Randall, D.A., Harshvardhan and D.A. Dazlich, 1991a: Diurnal New Zealand region inferred from general circulation models. variability of the hydrologic cycle in a general circulation New Zealand Meteorological Service, 142 pp. model. J. Atmos. Sci., 48, 40-61. Murphy, J.M., 1990: Prediction of the transient response of Randall, D.A., R.D. Cess, LP. Blanchet, G.J. Boer, D.A. climate to a gradual increase in CO2 using a coupled ocean/ Dazlich. A.D. Del Genio, M.Deque, V. Dymnikov, V. Galin, atmosphere model with flux correction. In: Research Activities S.J. Ghan, A.A. Lacis, H. Le Treut, Z.-X. Li, X.-Z. Liang, B.J, in Atmospheric and Oceanic Modelling. CAS/JSC Working McAvaney, V,P. Meleshko, J.F.B. Mitchell, J.-J. Morcrette. Group on Numerical Experimentation, Report No 14, G.L. Potter, L. Rikus, E. Roeckner, I.E. Royer, U. Schlese. WMO/TD No 396, 9.7-9.8. D.A. Sheinin, J. Slingo, A.P. Sokolov, K.E. Taylor, W.M. Nagai, T., T. Tokioka, M. Endoh and Y. Kitamura, 1991: El Washington, R.T. Wetherald, I. Yagai and M.-H. Zhang, Nino-Southern Oscillation simulated in an MRI atmosphere- 1991b: Intercomparison and interpretation of surface energy ocean coupled general circulation model. (Unpublished fluxes in atmospheric general circulation models. J. Geophys. manuscript) Res. (In press). Neelin, J.D., M. Latif, M.A.F. AOaart, M.A. Cane, U. Cubasch, Raper, S.B.C., T.M.L, Wigley and P.D. Jones, 1991: The effect W.L. Gates, P.R. Gent, M. Ghil, C. Gordon, N.-C. Lau, C.R. of man-made sulphate aerosols on empirical estimates of the Mechoso, G.A. Meehl, J.M. Oberhuber, S.G.H. Philander, P.S. climate sensitivity. (Unpublished manuscript). Schopf K.R. Sperber, A. Sterl, T. Tokioka, J. Tribbia and S.E. Rasch, P.J. and D.L. Williamson, 1990: Computational aspects Zebiak, 1992: Tropical air-sea interaction in general circulation of moisture transport in global models of the atmosphere. models. Climate Dyn. (In press). Quart. J. Roy. Met. Soc, 116, 1071-1090. Oberhuber, J.M., 1988: An atlas based on the COADS data set: Rasch, P.J. and D.L. Williamson, 1991: The sensitivity of a The budgets of heat, buoyancy and turbulent kinetic energy atgenera l circulation model climate to the moisture transport the surface of the global ocean. Max Planck Inst. Meteor., formulation. J. Geophys. Res., 96, 13123-13137. Report No. 15, Hamburg, 238 pp. Raval, A. and V. Ramanathan, 1989: Observational deter• Oglesby, R.J. and B. Saltzman, 1990: Sensitivity of the equi• mination of the greenhouse effect. Nature, 342, 758-761. librium surface temperature of a GCM to systematic changes Rind, D., 1991: Climate variability and climate change. In: in atmospheric carbon dioxide. Geophy. Res. Letters, 17, 1089- Greenhouse-Gas-Induced Climatic Change: A Critical 1092. Appraisal of Simulations and Observations, Elsevier, Oort, A.H., 1983: Global Atmospheric Circulation Statistics Amsterdam, pp. 69-78. 1958-1973. NOAA Professional Paper 14, U.S. Dept. Rind, D., E.-W. Chiou, W, Chu, J, Larsen, S. Oltmans, J. Lerner, Commerce, Washington, DC, USA. 180 pp. M.P. McCormick and L. McMaster, 1991: Positive water Owens, W.B. and P. Lemke, 1990: Sensitivity studies with a sea- vapor feedback in climate models confirmed by satellite data. ice mixed-layer pycnocline model in the Weddell Sea. J. Nature, 349, 500-503. Geophys. Res., 95, C6, 9527-9538. Robock, A., R.P. Turco, M.A. Harwell, T.P. Ackerman, R. Palmer, T.N., C. Brankovic, P. Viterbo and M.J. Miller, 1991: Andressen, H.-S. Chang and M.V.K., Sivakumar, 1991: Use of Modelling interannual variations of summer monsoons. J. general circulation model output in the creation of climate Clim. (In press). change scenarios for impact analysis. (Unpublished Park, C.K., D.M. Straus and K.M. Lau, 1990: An evaluation of manuscript). the structure of tropical intra-seasonal oscillations in three Santer, B.D., U. Cubasch, K. Hasslemann, W. Bruggemann, H. general circulation models. J. Met Soc. Japan, 68, 403-417. Hoeck, E. Maier-Reimer and U. Mikolajewicz, 1991: Philander, S.G.H., R.C. Pacanowski, N.-C. Lau and M.J. Nath, Selecting components of a greenhouse gas fingerprint. In: 1991: A simulation of the Southern Oscilladon with a global Proceedings of First Demetra meeting on Climate Variability atmospheric GCM coupled to a high-resolution tropical Pacific and Global Change, Chianciano Terme, Italy. 28 Oct - 3 Nov. ocean GCM. J. Clim. (In press). Schlesinger, M.E., X. Jiang and R.J. Charlson, 1991: Pierrehumbert, R.T., 1991: The mixing paradigm: Towards a Implication of anthropogenic atmospheric sulphate for the Langrangian view of the general circulation. Proceedings of sensitivity of the climate system. J.C. AUred.and A.S. Nichols First DEMETRA Meedng on Global Change, Chianciano (Eds.). Proceedings of Conf.Global Climate Change, Los Terme, Italy. (In press). Alamos, New Mexico. 21-24 October. B Climate Modelling, Climate Prediction & Model Validation 133 Semtner, A.J., Jr. and R.M. Chervin, 1988: A simulation of the Washington, W.M. and G.A. Meehl, 1989: Climate sensitivity global ocean circulation with resolved eddies. J. Geophys. due to increased CO2: Experiments with a coupled atmosphere Res.,n, 15502-15522. and ocean general circulation model. Climate Dyn., 4, 1-38. Semtner, A.J., Jr. and R.M. Chervin, 1991: Ocean general Washington, W.M. and G.A. Meehl, 1991: Characteristics of circulation from global eddy-resolving model. (Unpublished coupled atmosphere-ocean CO2 sensitivity experiments with manuscript). different ocean formulations. In: Greenhouse-Gas-Induced Senior, CA. and J.F.B. Mitchell, 1992a: COj and climate: The Climatic Change: A Critical Appraisal of Simulations and impact of cloud parametrization. (Unpublished manuscript). Observations. Elsevier, Amsterdam, pp. 79-110. Senior, CA. and J F B Mitchell, 1992b: The dependence of Washington, W.M. and G.A. Meehl, 1992: Greenhouse climate sensitivity on the horizontal resolution of a GCM. sensitivity experiments with penetrative cumulus convection (Unpublished manuscript). and tropical cirrus effects. (Unpublished manuscript). Shine, K.P. and A. Sinha, 1991: Sensitivity of the Earth's climate WCRP, 1988a: Validation of Satellite Precipitation Measure• to height dependent changes in the water vapour mixing ratio. ments for the Global Precipitation Climatology Project Nature, 354, 382-384. (Report of Intnl. Workshop, Washington, DC, Nov. 1986). Somerville, R.C.J, and L.A. Remer, 1984: Cloud optical thick• World Climate Research Programme, WCRP-1, WMO/TD - ness feedbacks in the CO2 problem. J. Geophys. Res., 89, No. 203, WMO, Geneva, 27pp. 9668-9672. WCRP, 1988b: World Ocean Circulation Experiment Imple• Stocker, T.F. and D.G, Wright, 1991: Rapid transitions of the mentation Plan. Vol. II, Scientific Background. World Climate ocean's deep circulation induced by changes in surface water Research Programme, WCRP-12, WMO/TD - No. 243, WMO, fluxes. Nature, 351, 729-732. Geneva, 132pp. Stouffer, R.J., S. Manabe and K. Bryan, 1989: Interhemispheric WCRP, 1990a: Report of the Fourth Session of the TOGA asymmetry in climate response to a gradual increase of Numerical Experimentation Group (Palisades, NY, USA, June atmospheric carbon dioxide. Nature, 342, 660-662. 1990). World Climate Research Programme, WCRP-50, Tibaldi, S. and F. Molteni, 1990: On the operational WMO/TD - No. 393, WMO, Geneva, 24pp. predictability of blocking. Tellus, 42A, 343-365. WCRP, 1990b: Report of the Second Session of the Monsoon Tibaldi, S., 1992: Low-frequency variability and blocking as Numerical Experimentation Group (Kona, HI, July 1990). diagnostic tools for global climate models. In Proceedings of World Climate Research Programme, WCRP-49, WMO/TD - the NATO Workshop on Prediction of Interannual Climate No. 392, WMO, Geneva, 16pp. Variations. 22-26 July, 1991, Trieste, Italy. J. Shukla (Ed.). WCRP, 1990c: Scientific Plan for the Global Energy and Water (Also appears in Proceedings of the Fondazione S. Paolo di Cycle Experiment. World Climate Research Programme, Torino Workshop on Oceans, Climate and Man. 15-17 April, WCRP-40 WMO/TD - No. 376, WMO, Geneva, 83pp. 1991, Turin, Italy.) WCRP, 1991: The Global Climate Observing System (Report of Toggweiler, J.R., K. Dixon and K. Bryan, 1989: Simulations of meeting, Winchester, UK, January 1991). World Climate radiocarbon in a coarse-resolution world ocean model. I. Research Programme, WCRP-56, WMO/TD - No. 412, WMO, Steady-state prebomb distributions. /. Geophys. Res., 94, Geneva, 15pp. 8217-8242. WCRP, 1992: The transient climate response to increasing Trenberth, K.E. and J.G. Olson, 1988: ECMWF global analyses greenhouse gases as simulated by four coupled atmosphere- 1979-1986: Circulation statistics and data evaluation. NCAR ocean GCMs. Report of the Steering Group on Global Climate Tech. Note TN -300 +STR, 94pp. Modelling (SGGCM), World Climate Research Programme, Tselioudis, G., W. Rossow and D. Rind, 1991: Global patterns of Geneva. (Unpublished manuscript). cloud optical thickness variation with temperature. Weaver, A.J. and E.S. Sarachik, 1991: The role of mixed (Unpublished manuscript). boundary conditions in numerical models of the ocean's Vinnikov, K. Ya. and I.B. Yeserkepova, 1991: Soil moisture: climate. J. Phys. Oceanogr (Special Edition on Modelling the Empirical data and model results. J. Climate, 4, 66-79. Large-scale Circulation of the Ocean). (In press). Von Storch, H., E. Zorita and U. Cubasch, 199\: Downscaling of Weaver, A.J., E.S. Sarachik and J. Marotzke, 1991: Internal low- global climate estimates to regional scales: An application to frequency variability of the ocean's thermohaline circulation. Iberian rainfall in wintertime. Max Planck Inst. Meteor. Nature. (In press). Report No. 64, Hamburg. Webb, D.J., P.D. Killworth, A.C. Coward and S.R. Thompson, Wang, H.J., Q.C Zeng and X-H. Zhang, 1991a: Simulation of 1991: The FRAM Atlas of the Southern Ocean. The Natural climate change due to doubled atmospheric CO2 using the lAP Environment Research Council, Swindon, UK. GCM with a mixed layer ocean. (Unpublished manuscript). Wigley, T.M.L., P.D. Jones, K.R. Briffa and G. Smith, 1990: Wang, W.-C, M.P. Dudek, X-Z. Liang and J.T. Kiehl, 1991b: Obtaining sub-grid-scale information from coarse-resolution Inadequacy of effective CO2 as a proxy in simulating the general circulation model output. /. Geophys. Res., 95, 1943- greenhouse effect of other radiatively active gases. Nature, 1953. 350, 573-577. Wigley, T.M.L. and S.C.B. Raper, 1990: Detection of the Wang, W.-C, M.P. Dudek and X-Z. Liang, 1992: A general enhanced greenhouse effect on climate. In: Climate Change: circulation model study of the climatic effect of atmospheric Science, Impacts and Policy. Proceedings Second World trace gases. (Unpublished manuscript). Climate Conference. Eds: J. Jager and H.L. Ferguson. Cambridge Univ. Press, UK. pp. 231-242. 134 Climate Modelling, Climate Prediction & Model Validation B Wright, P.B., 1988: An atlas based on the COADS data set: Fields of mean wind, cloudiness and humidity at the surface of the global ocean. Max Planck Inst. Meteor. Report No. 14, Hamburg. 70 pp. Zeng, Q.-C, G.C. Yuan, W.-Q. Wang and X.-H. Zhang. 1990: Experiments in numerical extra-seasonal prediction of climate anomalies. J. Chinese Acad. Sci., 14, 10-15.